The microbiological outcomes of culture-negative blood specimens using 16S rRNA broad-range PCR sequencing: a retrospective study in a Canadian province from 2018 to 2022

ABSTRACT

Broad-range 16S rRNA PCR and sequencing of 1,183 blood specimens from 853 unique patients yielded an interpretable sequence and bacterial identification in 29%, 16S rRNA amplification with uninterpretable sequences in 53%, and no amplification in 18%. This study highlights the potential utility of this technique in identifying fastidious gram-negative and anaerobic bacteria but the frequent recovery of environmental and contaminant organisms argues for its judicious use.

IMPORTANCE

The existing literature focuses on its performance compared to blood cultures in patients with sepsis, leaving a gap in the literature regarding other blood specimens in suspected infectious syndrome across the severity spectrum. We aimed to characterize its microbiological outcomes and provide insight into its potential clinical utility.

INTRODUCTION

The accurate identification of bacterial pathogens is essential for tailored antimicrobial therapy. In suspected infectious syndromes, the inability to identify the causative organisms can lead to prolonged use of broad-spectrum antibiotics with increased risk of adverse effects, selective pressure leading to bacterial resistance, and Clostridioides difficile infection (1).

Despite recent technological advancements, the primary method of bacterial identification in clinical specimens remains traditional culture-based approaches. However, this approach has well-described limitations, including false-negative results from prior administration of antibiotics, insufficient blood culture volumes, and the limited capacity to recover fastidious organisms using routine media (2, 3).

By contrast, nucleic acid amplification tests (NAATs) are not reliant on the viability of organisms and have demonstrated utility in the above clinical scenarios (4–6). Among these techniques, broad-range bacterial PCR sequencing (BRBPS) of the 16S rRNA gene, a highly conserved locus, has emerged as a potentially valuable diagnostic tool (6). While its clinical utility has been characterized for different specimen types, including cardiac valve tissue, cerebrospinal fluid, and musculoskeletal tissue, its role in blood specimens is not well defined. Particularly with EDTA blood specimens and specimens incubated in automated blood culture systems that failed to yield growth on subculture, its utility remains incompletely characterized (6). The existing literature focuses on its performance compared to blood cultures in patients with sepsis, leaving a gap in the literature regarding other blood specimens in suspected infectious syndrome across the severity spectrum (4, 7). To address this, we conducted a detailed retrospective microbiological analysis of all BRBPS of blood specimens performed in the Laboratoire de santé publique du Québec (LSPQ), the provincial reference laboratory for the Province of Québec, Canada. We aimed to characterize its microbiological outcomes and provide insight into its potential clinical utility.

METHODS

Study design and data collection

We completed a retrospective study at the LSPQ from June 11, 2018 to October 31, 2022. The LSPQ provides exclusive 16S BRBPS testing for a population of over 8.6 million. We reviewed all clinical blood specimens sent for BRBPS, including EDTA blood, serum, and blood culture bottles. For incubated blood culture bottles, testing was performed on bottles where no growth on subculture was identified by the referring laboratory, regardless of the Gram stain result. Details on the specific type of blood culture bottles (e.g., aerobic, anaerobic, or pediatric) and culture conditions were not available.

The primary analysis was restricted to the first specimen obtained for each patient. Microbiological outcomes were categorized into three categories: interpretable sequence, uninterpretable sequence, or negative PCR result. Uninterpretable sequences were those with successful PCR amplification but with Sanger sequencing reactions that were either too short or with a mixed chromatogram. Interpretable sequences were identified to the species or genus level, as below, and subsequently classified using the Centres for Disease Control National Healthcare Safety Network database (NHSN; v9.2 2019). Sequences classified as possible pathogens and organisms associated with mucosal barrier injury (MBI) according to NHSN were considered potentially clinically significant, while common commensals were considered possible contaminants. We pragmatically subdivided the first two groups into the broad categories used in medical microbiology (fastidious gram-negative bacilli, anaerobes, gram-positive bacilli, gram-positive cocci, Enterobacterales, non-fermenter gram-negative bacilli, and “environmental organisms”). We defined environmental bacteria as those predominantly found in environmental reservoirs with low pathogenicity to humans, and being extremely unlikely to cause bacteremia. The data were analyzed using standard descriptive statistics. Categorical data such as frequencies were compared using the chi-square test. A P-value of <0.05 was considered statistically significant. SPSS version 25 (IBM, Chicago, Illinois) was used to perform the analyses.

This study was conducted as a laboratory improvement method. De-identification was made at the reference laboratory prior to the study onset. At the provincial reference laboratory (LSPQ), a diagnostic test is performed after informed consent from patients, and the project was considered exempt from IRB approval when its objective is to improve diagnostic methods.

16S PCR methodology

All specimens (200 µL) were treated with proteinase K (10 minutes) at room temperature and then extracted with QIAsymphony (Hilden, Germany) according to the manufacturer’s instructions. A negative and a positive control (Ochrobactrum anthropi) were included for each extraction. A portion of the 16S rRNA gene was amplified by PCR using primers 8–27F (5′-AGAGTTTGATCMTGGCTCAG-3′) and 803R (5′-CTACCAGGGTATCTAATCC-3′) (8, 9). The 796 bp amplified fragment was purified using the MinElute filtration system QIAGEN (Hilden, Germany) and sequenced on both strands using the 8–27F, 803R, and 503R (5′-TTACCGCGGCTGCTGG-3′) (7, 10). Sequences were determined with an ABI 3500 XL sequencer using a BigDye sequencing kit V3.1 (Applied Biosystems). Consensus sequences of ≥350 base pairs were used for identification. The sequences were subjected to a BLAST search followed by multiple sequence alignments using closely related sequences with the ClustalW program (10, 11). Phylogenetic analysis was performed using the Bionumerics V7.6.3 (Applied Maths NV, Sint-Martens-Latem, Belgium). Final identification was made by comparing closely related sequences from NCBI and LSPQ databases.

RESULTS

A total of 1,183 blood specimens from 853 unique patient identifier numbers underwent BRBPS. The primary analysis included 567 blood culture bottles, 196 EDTA blood, 8 serum, and 82 unspecified blood specimens (Table 1). Specimens originated from males more than females (57.8%), with an average age of 60.7 years (SD 20.0). PCR amplification failed in 18.3% (n = 156) of specimens, yielded uninterpretable sequences in 52.7% (n = 450), and resulted in an interpretable sequence in 247 (29.0%) specimens (Fig. 1). Of those with interpretable sequences, we identified 172 specimens with potentially clinically significant results, including MBI organisms (n = 88) and possible pathogens (n = 84), as shown in Fig. 2. In all, 44 patients (17.8%) had possible contaminants identified. Most identified potentially clinically significant organisms were classified as anaerobes (n = 68, 39.3%) or fastidious gram-negative bacteria (n = 46, 26.6%). The latter consisted of 30 (65.2%) zoonotic-associated bacteria and 5 (10.9%) arthropod-associated bacteria. Environmental bacteria accounted for (n = 15) 17.9% of possible pathogens.

TABLE 1.

Basic demographics and microbiological characteristics of BRBPS of the 16S rRNA on culture-negative blood specimens

	All specimens n = 853	16S PCR negative n = 156 (18.3)	Uninterpretable sequences n = 450 (52.7)	Interpretable sequences n = 247 (29.0)
Age, mean (SD)	60.7 (20.0)	59.6 (19.2)	62.2 (19.3)	58.7 (21.6)
Sex, % male	57.8	59.0	57.7	57.3
Specimen condition, n (%)
Refrigerated	103 (12.1)	17 (16.5)	50 (48.5)	36 (35.0)
Frozen	552 (64.7)	108 (19.5)	320 (58.0)	124 (22.5)
Ambient	130 (15.2)	21 (16.2)	54 (41.5)	55 (42.3)
Not available	68 (8.0)	10 (14.7)	26 (38.2)	32 (47.1)
Specimen, n (%)
Blood culture bottles	567 (66.5)	59 (10.4)	289 (51.0)	219 (38.6)
Refrigerated	63 (11.1)	6 (9.5)	28 (44.4)	29 (46.0)
Frozen	325 (57.3)	34 (9.8)	184 (56.6)	107 (32.9)
Ambient	116 (20.5)	11 (9.4)	52 (44.8)	53 (45.7)
Not Available	63 (11.1)	8 (12.6)	26 (41.3)	25 (39.7)
EDTA blood	196 (22.9)	36 (18.4)	148 (75.5)	12 (6.1)
Serum	8 (0.9)	2 (25.0)	4 (50.0)	2 (25.0)
Blood, NOS	82 (9.6)	59 (72.0)	8 (9.7)	15 (18.3)
Gram stain of incubated blood cultures bottle, n (%)
Positive	67 (11.8)	3 (4.5)	10 (14.9)	54 (80.6)
Negativea	2 (0.0)	0	1 (50.0)	1 (50.0)
Not available	498 (88.2)	56 (11.2)	278 (55.8)	164 (33.0)

^a Negative on Gram stain, but positive on acridine orange. Abbreviations: BRBPS, broad-range bacterial polymerase chain reaction sequencing; SD standard deviation; EDTA, ethylenediaminetetraacetic acid; NOS, not otherwise specified.

Fig 1.

Study design flow diagram summarizing the microbiological outcome of specimens. Abbreviations: NHSN, National Healthcare Safety Network; NS, non-specific.

Fig 2.

Treemap of the potential pathogens identified in blood specimens by 16S gene sequencing, organized by microbiological characteristics. The proportions of the organism identified by 16S gene sequences in the (A) potential pathogen and (B) mucosal barrier injury categories by the relative area of the colored polygons. Organism categories are based on the National Healthcare Safety Network database. The Treemap is subcategorized by microbiological characteristics: pink: Fastidious gram-negative, gray: anaerobes, blue: environmental, green: non-fermenter gram-negative bacilli, yellow: gram-positive cocci, purple: gram-positive bacilli, orange: Enterobacterales, and brown: Other.

Blood specimens received at room temperature (55/130, 42.3%) had a significantly higher yield (P < 0.01) of an interpretable sequence relative to frozen (124/552, 22.5%) or refrigerated specimens (36/103, 35.0%) (Table 1). When restricted to blood culture specimens, the yield of interpretable sequences for ambient (53/116, 45.7%) and refrigerated specimens (29/63, 46.0%) were significantly higher than frozen blood cultures (107/325, 32.9%) (P < 0.01). Furthermore, blood culture bottle specimens had a higher proportion of interpretable sequences than EDTA blood specimens (219/567, 38.6% vs 12/196, 6.1%) (P < 0.01). Gram stain results were infrequently recorded on the requisition forms. However, in cases where a positive result was documented by the referring laboratory (67/567, 11.8%), the proportion with an interpretable sequence was 80.6%, of which 53.7% were identified as anaerobes and 24.1% as fastidious gram-negative bacteria. In addition, we observed a high concordance of 79.6% (43/54) between interpretable sequences and Gram stain results.

Excluded from the primary analysis were 330 specimens corresponding to 174 unique patient identifier numbers who underwent multiple BRBPS testing within a 31-day period with a median time interval of 0 days (IQR 0–2) between tests. Among these, repeated tests had limited clinical utility in 156 (89.7%) instances; 116 had concordant results, and 40 had possible contaminants detected in at least one sample. By contrast, discordant results were observed in 18 (10.3%) instances with possible pathogens or MBI organisms identified in at least one of the specimens.

DISCUSSION

In this retrospective study, the overall interpretable sequence yield of 16S rRNA BRBPS in blood specimens, including blood cultures, and EDTA blood or serum specimens, was 29%. However, this decreased to 20% when limited to potentially clinically significant results. Fastidious gram-negative bacteria associated with zoonotic or arthropod exposures and anaerobic bacteria were overrepresented. This suggests an important potential clinical utility of BRBPS in blood specimens and specific scenarios in culture-negative infectious syndromes where these bacteria are suspected. However, the frequent recovery of possible contaminants and environmental organisms with this technology highlights the importance of laboratory stewardship to ensure its appropriate use. In addition, to be clinically meaningful, the assay’s turnaround time must be less than the duration of empiric antimicrobial therapy for relevant infectious syndromes (<6 weeks).

Although data were limited, when reported, culture-negative endovascular infection constituted the most common clinical indication. The recently updated DUKE-ISCVID diagnostic criteria for infective endocarditis include molecular diagnostic methods, including BRBPS for blood specimens. Specifically, a positive result for Bartonella spp., Coxiella burnetii, or Tropheryma whipplei now constitutes a major criterion (12). Only two Bartonella spp and no cases of C. burnetii or T. whipplei were identified over this period while the LSPQ has identified eight cases of C. burnetii from tissue (author’s unpublished data). In addition, only a small proportion (2.3%) of test results identified typical infective endocarditis organisms for native valve endocarditis, which are generally culture positive and thus not submitted for BRBPS. These findings suggest that BRBPS of blood specimens may not be a high-yield tool for diagnosing infective endocarditis in patients with suspected culture-negative endocarditis. However, additional clinical data would be required to assess this thoroughly.

In instances where blood culture specimens were positive by Gram stain result but failed to grow on subculture, BRBPS successfully identified a pathogen in approximately 80% of cases. The identified organisms were primarily fastidious, comprising mainly anaerobes and fastidious gram-negative bacilli, accounting for the lack of growth on primary media. This finding highlights the potential utility of BRBPS in the diagnostic workflow for blood culture specimens with suspected fastidious bacteria. In addition, we identified specific preanalytical characteristics associated with improved test performance. First, as stated, specimens obtained from automated blood culture systems demonstrated a higher interpretable sequence yield, likely attributed to the increased biological amplification during incubation and, therefore, more bacterial DNA pre-BRBPS. Conversely, frozen specimens had lower recovery yield, potentially due to DNA degradation from thawing-freezing cycles (13). Overall, incubated blood cultures sent at room temperature or refrigerated are preferred to maximize the interpretable sequence yield of an organism with BRBPS. Analysis of repeated testing within a 31-day period has limited utility, as the majority of results were concordant or detected possible contaminants.

Our results must be interpreted in the context of the study characteristics. The clinical data available to correlate with our microbiological results were restricted to requisition forms and limited our ability to assess the clinical utility of BRBPS at the individual patient level. While our study mainly involved patients with suspected culture-negative bloodstream infections, these findings may not be generalizable to other clinical syndromes. There was also a significant proportion of results yielding uninterpretable sequences despite successful PCR amplification of the 16S rRNA gene. The Sanger sequencing method frequently cannot produce a consensus sequence in the presence of mixed amplified PCR products. To overcome this limitation, further technical optimization, including targeted or untargeted next-generation sequencing (NGS) approaches, is required to improve the interpretable sequence yield as NGS-based workflows can discriminate mixed sequences (7).

In summary, BRBPS on blood specimens may be helpful in specific clinical scenarios, for example, when zoonotic or arthropod-associated and anaerobic bacteria are suspected. This is particularly true in the case of fastidious organisms suspected of the failure of an incubated blood culture bottle to grow on subculture. However, considering that a significant proportion of PCR specimens produced uninterpretable results, and a notable proportion were possible contaminants or environmental organisms, these factors argue for careful and judicious use.