Plasma Microbial Cell-Free DNA Sequencing for Pathogen Detection and Quantification in Children With Musculoskeletal Infections

Abstract

Background

Nearly half of all pediatric musculoskeletal infections (MSKIs) are culture negative. Plasma microbial cell-free DNA (mcfDNA) sequencing is noninvasive and not prone to the barriers of culture. We evaluated the performance of plasma mcfDNA sequencing in identifying a pathogen, and examined the duration of pathogen detection in children with MSKIs.

Methods

We conducted a prospective study of children, aged 6 months to 18 years, hospitalized from July 2019 to May 2022 with MSKIs, in whom we obtained serial plasma mcfDNA sequencing samples and compared the results with cultures.

Results

A pathogen was recovered by culture in 23 of 34 (68%) participants, and by initial mcfDNA sequencing in 25 of 31 (81%) participants. Multiple pathogens were detected in the majority (56%) of positive initial samples. Complete concordance with culture (all organisms accounted for by both methods) was 32%, partial concordance (at least one of the same organism(s) identified by both methods) was 36%, and discordance was 32%. mcfDNA sequencing was more likely to show concordance (complete or partial) if obtained prior to a surgical procedure (82%), compared with after (20%), (RR 4.12 [95% CI 1.25, 22.93], p = .02). There was no difference in concordance based on timing of antibiotics (presample antibiotics 60% vs no antibiotics 75%, RR 0.8 [95% CI 0.40, 1.46], p = .65]). mcfDNA sequencing was positive in 67% of culture-negative infections and detected a pathogen for a longer interval than blood culture (median 2 days [IQR 1, 6 days] vs 1 day [1, 1 day], p < .01).

Conclusions

Plasma mcfDNA sequencing may be useful in culture-negative pediatric MSKIs if the sample is obtained prior to surgery. However, results must be interpreted in the appropriate clinical context as multiple pathogens are frequently detected supporting the need for diagnostic stewardship.

INTRODUCTION

Osteomyelitis and septic arthritis, together known as musculoskeletal infections (MSKIs), have the potential to cause serious acute and chronic complications [1]. To achieve optimal outcomes, early diagnosis and targeted treatment are essential [1, 2]. Targeted antibiotic therapy requires the identification of a pathogen by culture or molecular techniques from a sterile site, often necessitating an invasive procedure (eg, bone biopsy). In nearly half of all pediatric MSKIs, no organism is identified [3–6]. Several reasons for this exist, including antibiotic use prior to obtaining a culture; lack of a surgical procedure; and the fastidious nature of certain organisms in culture [7]. Absence of a microbiologic diagnosis has significant ramifications, as patients are treated with broad, often multi-agent antimicrobial regimens. The overuse of broad-spectrum antibiotics is associated with an increased risk of adverse events, and increases the risk of developing antimicrobial resistance (AMR) [8–13]. Improved culture-independent pathogen diagnostic technologies are needed.

Plasma microbial cell-free DNA (mcfDNA) sequencing is one such culture-independent diagnostic technology with the potential to improve pathogen identification for children with MSKIs. This pathogen-agnostic technology utilizes detection of circulating mcfDNA to identify and quantify pathogens [14, 15], is noninvasive and not prone to the barriers typically associated with pathogen identification in children with MSKIs.

In this study, we sought to compare results of plasma mcfDNA sequencing with standard culture-based diagnostics in identifying a pathogen in children with MSKIs. Secondarily, we examined the duration of pathogen detection by plasma mcfDNA sequencing over the course of treatment in children with MSKIs.

METHODS

Patient Enrollment

We prospectively enrolled children 6 months to 18 years of age admitted to Riley Hospital for Children (Indianapolis, IN), from July 2019 to May 2022. Potential study participants were identified by daily screening of the pediatric infectious diseases and hospitalist patient censuses, and approached for enrollment if there was a strong clinical suspicion of MSKI as evidenced by fever, musculoskeletal pain (eg, tenderness to palpation of a joint/bone, or refusal to bear weight), and elevated C-reactive protein (CRP) or erythrocyte sedimentation rate (ESR). Informed consent was obtained from a guardian, and informed assent from the participant when able, before inclusion in the study. Patients were excluded if there was clinical evidence suggesting an alternative diagnosis at the time of enrollment or patient/guardian was unable or unwilling to consent for the study. The plasma mcfDNA sequencing results were not released to the clinical team. The study was approved by the Indiana University School of Medicine Institutional Review Board.

Specimen Collection

Study sample collection is outlined in Supplementary Figure 1. Serial plasma samples were obtained, when possible, during inpatient and outpatient treatment. The initial study sample was obtained within 96 h of presentation to the hospital, and as close to the initial blood culture as possible, but often not during the same blood draw. A second sample was obtained 12–72 h after the initial sample. Together these samples were called “detection samples.” Detection samples were sent to Karius, Inc. (Redwood City, CA) for real-time testing. If either detection sample was positive for an organism, participants were eligible to have subsequent plasma mcfDNA sequencing samples drawn. Subsequent samples were obtained up to every 48 h while the participant was hospitalized (maximum of two subsequent inpatient samples, total of four inpatient samples), followed by outpatient samples obtained during routine clinic visits (maximum of three). Subsequent samples were stored at −70°C, and sent to Karius, Inc. for batch testing. All plasma mcfDNA sequencing samples were obtained during routine blood draws or procedures as part of clinical care. If a participant did not have a blood draw/procedure, was discharged from the hospital (inpatient samples only), or did not have an outpatient clinic appointment at the time of a study sample collection, the study sample was not obtained.

Specimen Processing for mcfDNA Sequence Testing

Whole blood was collected in EDTA-containing tubes and centrifuged, within 6 h of blood draw, at 1,100 RCF for 10 min at room temperature. Plasma was transferred into sterile polypropylene tubes. Detection samples were sent at ambient temperature to Karius, Inc. within 96 h of the blood draw. Subsequent samples were stored at −70°C until they were shipped, on dry ice, to Karius, Inc. All samples were de-identified, and Karius, Inc. was blinded to culture results.

Once received at Karius, Inc., samples were processed as follows: cfDNA was extracted, converted to DNA libraries, sequenced using Illumina NextSeq 500, with removal of human sequences [14, 15]. Once sequenced, the results were run through the Karius bioinformatics pipeline (version 3.13), which is responsible for mapping sequencing data to a curated database of genomic references, estimating the concentration for each of the detected microbes, and the applications of a series of quality control tests [14, 15]. Organisms identified were quantified in molecules per microliter (MPM) which represents the number of DNA sequencing reads present per microliter of plasma. Statistically significant levels of mcfDNA above background were reported as a positive result.

Data Analysis

Demographic and clinical data, for up to 1 year after hospitalization, was recorded from the electronic medical record. Descriptive statistics were used to describe the study population. Initial plasma mcfDNA sequencing results were compared with initial culture results (blood and surgical) for the primary analysis, and subset analysis was done by culture positivity type (blood vs surgical), infection type (osteomyelitis vs septic arthritis vs both), and antibiotic exposure (before or after plasma mcfDNA sequencing sample). Complete concordance was defined as the same organism(s) identified by both culture and mcfDNA sequencing (all organisms accounted for in both samples). Partial concordance was defined as at least one of the same organisms identified in culture and mcfDNA sequencing, but not all organisms accounted for in both samples. Organisms in the culture that were considered contaminants by the clinical team were treated as a contaminant for comparison purposes, and not counted as a positive culture. Univariate analysis using Fisher’s exact test was used to compare concordance among culture and plasma mcfDNA sequencing results between groups of interest (surgical procedure prior to study sample vs no procedure; antibiotic pretreatment prior to study sample vs no antibiotic pretreatment). For secondary outcomes, we assessed test duration positivity by comparing the percent of participants with a positive test (plasma mcfDNA sequencing vs blood culture) using survival curves (generated using GraphPad Prism; San Diego, CA). Duration of positivity curves for each test method (blood culture vs mcfDNA sequencing) was compared using a log-rank (Mantel–Cox) test.

RESULTS

Patient Characteristics

During the study period, 42 patients were approached for enrollment. Five parents/guardians were unwilling/unable to consent for the study, 2 participants ultimately had an alternative diagnosis (transient synovitis) and 1 patient was enrolled but did not have study labs drawn, resulting in 34 participants for the final analysis. Demographic and clinical characteristics are summarized in Table 1. The majority of participants were non-Hispanic white (74%), male (71%), and without a preexisting medical condition (79%). The median age was 8 (IQR 4–12) years. Osteomyelitis was the most common infection type (62%), followed by septic arthritis (24%), and concurrent osteomyelitis and septic arthritis (15%). Most participants had at least one surgical procedure (71%). An initial plasma mcfDNA sequencing sample was drawn in 32 participants (2 had subsequent, but not initial samples drawn) and 1 failed quality control leaving 31 participants for the primary analysis. Of those, 16 had an initial plasma mcfDNA sequencing sample drawn within 24 h of the initial blood culture (3 during the same blood draw). The median time between the initial blood culture and plasma mcfDNA sequencing sample was 19 h (IQR 4, 31 h).

Pathogen Identification by Culture

A pathogen was recovered by culture in 23 of 34 participants (68%). Cultures were positive by blood culture only in seven participants (21%), by surgical culture only in eight participants (24%), and in both blood and surgical cultures in eight participants (24%). Overall, 15 of 34 (44%) of participants had a positive blood culture (Table 2).

Pathogen Identification by Plasma mcfDNA Sequencing

The initial plasma mcfDNA sequencing sample was positive in 25 of 31 (81%) of the participants (1 sample failed quality control and 2 participants did not have an initial inpatient study sample drawn, but had subsequent samples drawn). In 11 of 25 (44%), a single organism was detected by the plasma mcfDNA sequencing and in 14 of 25 (56%) multiple organisms were detected.

The initial plasma mcfDNA sequencing sample identified the same organism as culture (complete or partial concordance) in 15 of 22 (68%) of the participants with a positive culture (1 sample failed quality control; Table 2). Of these 15 samples, there was complete concordance in 7 and partial concordance in 8. Of the seven discordant results, plasma mcfDNA sequencing was negative in three, and found a different organism(s) in four. In two participants with a discordant initial sample, the second “detection sample” was concordant with culture (Table 3).

In subset analysis by culture type, the initial plasma mcfDNA sequencing sample identified the same organism as culture in 13 of 15 (87%) participants when the blood culture was positive. Of those 13, there was complete concordance in 5 and partial concordance in 8. The two mcfDNA sequencing samples that were discordant with blood cultures were drawn 54 and 87 h after the corresponding blood cultures, respectively. Plasma mcfDNA sequencing was concordant with culture in 2 of 7 (29%) when only the surgical culture was positive.

Analysis by infection type showed plasma mcfDNA sequencing identified a pathogen found in culture in 9 of 13 (69%, 4 complete concordance, 5 partial concordance) of participants with osteomyelitis, 4 of 4 (100%, all complete concordance) with septic arthritis, and 2 of 5 (40%, 1 complete concordance, 1 partial concordance) with concurrent osteomyelitis and septic arthritis.

When looking at plasma mcfDNA sequencing results based on timing of sample acquisition, plasma mcfDNA sequencing was more likely to identify the same organism as culture in those who did not have a surgical procedure prior to obtaining the study sample, 14 of 17 (82%), compared with those who had a surgical procedure prior to the study sample, 1 of 5 (20%), (RR 4.12 [95% CI 1.25, 22.93], p = .02). There was no difference in concordance between those who received antibiotic treatment prior to the study sample, 6 of 10 (60%), compared to those who did not, 9 of 12 (75%), (RR 0.8 [95% CI 0.40, 1.46], p = .65).

Of the 11 participants with negative cultures, 9 had an initial plasma mcfDNA sample drawn. Plasma mcfDNA sequencing was positive in six of nine (66%) culture-negative infections. Of the six positive plasma mcfDNA sequencing samples, Haemophilus influenzae was identified most commonly (four participants). Kingella kingae, was identified in one participant. Staphylococcus aureus was not identified by plasma mcfDNA sequencing in any culture-negative infections. Detailed comparisons of pathogen recovery by culture and plasma mcfDNA sequencing are shown in Table 3.

Serial Plasma mcfDNA Sequence Testing

Of the 15 participants in whom the initial plasma mcfDNA sequencing was concordant with culture (either complete or partial), 10 had a second plasma mcfDNA sequencing test sent within 48 h of the initial test. Plasma mcfDNA sequencing demonstrated persistent microbial detection in 7 of 10 (70%). Eight subsequent inpatient samples (in six participants) were drawn between 3 and 14 days of the initial sample. Plasma mcfDNA sequencing continued to detect a microbe in six of eight (75%). Additionally, 11 outpatient samples (in 8 participants) were sent between 8 and 48 days of the initial sample. Plasma mcfDNA sequencing continued to detect a microbe in 1 of 11 (9%). Overall, plasma mcfDNA sequencing detected a pathogen for a longer duration (median 2 days [IQR 1, 6 days]) compared with blood culture (median 1 day [IQR 1, 1 day], p < .01 by log-rank test; Figure 1).

Supplementary Table 1 shows the serial quantitative MPM levels of those participants with initial concordant results and serial samples (n = 13), compared to CRP, a commonly obtained inflammatory biomarker used as an indirect measure of clinical improvement for children with MSKIs.

DISCUSSION

This single-center, prospective study provides insights into the utility, as well as the limitations, of plasma-based mcfDNA sequencing for pathogen identification in children with MSKIs. We found that plasma mcfDNA sequencing identified an organism in a higher proportion of patients compared with culture, showed good concordance with culture when drawn prior to a surgical procedure and continued to detect a pathogen in the majority of hospitalized patients. However, plasma mcfDNA sequencing was less reliable when drawn after a surgical procedure, and frequently identifies multiple organisms. These results highlight the need for proper diagnostic stewardship and appropriate clinical context when obtaining plasma mcfDNA sequencing in children with MSKIs to maximize diagnostic yield and ensure results are accurately interpreted.

The potential for plasma mcfDNA sequence testing to revolutionize infectious disease diagnostics is well recognized [16]. As a plasma-based test with the potential to identify a broad range of pathogens, through a nontargeted approach, mcfDNA sequencing offers the ability to noninvasively identify the causative pathogen in a wide array of invasive infections, and is not limited by the barriers of traditional culture-based techniques. It has shown utility in infections of immunocompromised patients, in whom the list of potential pathogens is extensive, including neutropenic fever [17, 18] and respiratory tract infections [19], as well as in immunocompetent hosts with endocarditis [20], and focal central nervous system infections [21, 22].

While there is some data for the use of mcfDNA sequence testing in adults with MSKIs, particularly prosthetic joint infections [23–25], data in children are lacking. Given the significant differences in infection pathogenesis, with the majority of pediatric infections being due to acute hematogenous spread compared to direct inoculation or vascular insufficiency in adults, studying the performance of mcfDNA sequence testing in children with MSKIs is essential to optimize use in this population. In a pediatric study of children with MSKIs comparing culture and culture plus polymerase chain reaction (PCR) with metagenomic next-generation sequence (mNGS) testing from operative cultures, the investigators found that mNGS testing identified a pathogen in 61.9% of subjects, the diagnostic yield of mNGS testing was similar to culture/PCR, and concluded that mNGS testing did not significantly impact clinical care [26].

In our study, we found that plasma mcfDNA sequencing identified an organism more frequently than culture-based techniques (81% vs 68%, respectively). However, the clinical significance of many of the organisms identified is unclear. In culture-positive infections, concordance (partial or complete) with culture was only 68%. In culture-negative infections (the likely scenario in which a plasma mcfDNA sequencing test would be sent), an organism was identified in 67% of cases. Whether the organisms identified were the causative pathogen is unclear. Interestingly, we found that Haemophilus influenzae was a common organism detected by plasma mcfDNA sequencing (n = 11), either by itself (n = 4) or with other organisms (n = 7). H. influenzae was found in both culture-positive infections (n = 7, concordant with culture once) and culture-negative infections (n = 4). In the post-Hib vaccine era, H. influenzae is a relatively uncommon cause of MSKI [6, 27], though it does account for a small percentage of pediatric MSKI, including in 1 patient in our study. These findings need to be considered in clinical practice if H. influenzae is identified by plasma mcfDNA in children with MKSI, as it may not represent a true pathogen, thus narrowing antibiotics to target H. influenzae should be done with caution. Importantly, one case of K. kingae, a well-reported cause of MKSIs that is notoriously difficult to culture [28], was identified. These findings highlight both the challenges, as well as the potential benefits, of using plasma mcfDNA sequencing in children with MSKIs when cultures are negative.

Importantly, our study suggests that timing of sample acquisition is important. mcfDNA sequencing identified the same organism as culture (partial or complete concordance) in 82% (14 of 17) when a sample was obtained prior to a surgical procedure (though complete concordance was seen in only 35%), and dropped to 20% (1 of 5) when obtained after a surgical procedure. In the four discordant cases, the samples were obtained 16, 18, 21, and 52 h after the surgical procedure, respectively. This suggests that these infections often can be cleared quickly from the blood with adequate source control. The ability of plasma mcfDNA sequencing to identify a pathogen relies on circulating pathogen cell-free DNA in the blood, and our findings suggest that once a surgical procedure has been performed, the level of mcfDNA decreases substantially, thus limiting the utility of the test. In clinical practice, if a plasma mcfDNA sequencing test is being considered, our study suggests the sample should be drawn prior to a surgical procedure and sent for testing or frozen and sent for testing if cultures remain negative.

Our study also highlighted a known drawback of plasma mcfDNA sequencing, finding that multiple organisms were identified in the majority (56%) of participants with a positive sample. This is well reported in studies investigating the use of plasma mcfDNA sequencing in other infection types [29, 30]. The hypothesized pathogenesis of acute hematogenous MSKIs (the mechanism in the majority of participants in this study) is that bacteria colonize the nose and oropharynx, breach the mucosal barrier, travel through the bloodstream, and establish infection in the bone or joint [31]. The pathogenesis of these infections may explain why the multiple pathogens, many of which are common oropharyngeal colonizers, were identified in the majority of participants, as more than one organism may breach the mucosal barrier, and thus be detected by nontargeted plasma mcfDNA sequencing. These results highlight the importance of having experienced clinicians, comfortable with interpreting plasma mcfDNA sequencing results, involved in the care of these patients.

With the ability to provide a quantitative result (molecules per microliter), it has also been proposed to use plasma mcfDNA sequencing serially as a biomarker to assess clinical improvement. We found that plasma mcfDNA sequencing continued to detect a pathogen in the majority of hospitalized patients who had serial inpatient testing. Additionally, we found that serial MPMs followed a similar trend as CRP, an inflammatory marker commonly followed in children with acute MSKI. Appropriately powered studies to assess the use of quantitative mcfDNA as a clinical biomarker of improvement are warranted.

Despite the strengths of our study, there are several limitations. First, the relatively small sample size limits the power of the results, especially subset analysis such as by infection type or by pathogen. Another limitation of the study is the use of culture as a gold standard for which to compare plasma mcfDNA sequencing. This study was designed to compare plasma mcfDNA sequencing with standard-of-care testing, conventional culture or molecular, at the discretion of the clinical team. For all participants, conventional culture was the comparator (no molecular tests, eg, PCR, were performed by the clinical team). This limits the ability to identify fastidious organisms, particularly K. kingae, a common pathogen causing MSKIs in children <4 years of age. Additionally, the initial plasma mcfDNA sequencing tests, as well as subsequent tests, were obtained with standard-of-care laboratory samples, resulting in the plasma mcfDNA sequencing samples being obtained at various time points during hospitalization. Although this likely reflects the real-world use of the test, having standardized time points would decrease the variability between samples and provide a more consistent assessment of the performance of the test at each time point. Similarly, obtaining plasma mcfDNA sequencing samples at the same time as blood culture would have allowed for a more direct comparison of the two tests, but the study design (blood cultures often drawn in the emergency room prior to enrollment) did not allow for this. Also, since investigators were blinded to the plasma mcfDNA sequencing results the study could not evaluate the impact of the test on patient management. Finally, at the time of this study, AMR marker detection (eg, mecA) via plasma mcfDNA sequencing was not available. These capabilities are now available, and analysis to assess the accuracy of AMR marker detection is planned.

CONCLUSIONS

Increasing rates of AMR have necessitated the development of improved, culture-independent infectious disease diagnostic tests. Plasma mcfDNA sequencing provides clinicians with a noninvasive tool with the potential to detect a broad range of pathogens. In our study of children with MSKIs, we found that plasma mcfDNA sequencing identified an organism in a higher proportion of cases compared with culture and was reliable when obtained prior to a surgical procedure, highlighting its potential utility in culture-negative infections if obtained prior to a surgical procedure. However, our study also highlights the limitations of plasma mcfDNA sequencing in children with MSKIs, namely that reliability decreases when obtained after a surgical procedure and that results may have unclear clinical relevance as multiple organisms are frequently identified. Taken together, plasma mcfDNA may have utility in culture-negative pediatric MSKIs, however, diagnostic stewardship is essential to maximize yield, and results must be interpreted in the appropriate clinical context, ideally by experienced clinicians, comfortable with interpreting plasma mcfDNA sequencing results.

Financial Support.

This study was supported by Karius Inc. Employees of Karius Inc are named as authors on this manuscript and contributed to the study design and preparation of the manuscript.

Potential conflicts of interest.

J.W. and J.S. received grant support to their institutuion from Karius Inc. M.S. and S.P. are employees of Karius Inc.

REFERENCES