Cell‐free DNA next‐generation sequencing successfully detects infectious pathogens in pediatric oncology and hematopoietic stem cell transplant patients at risk for invasive fungal disease

We sought to determine if next‐generation sequencing (NGS) of microbial cell‐free DNA (cfDNA) in plasma would detect pathogens in pediatric patients at risk for invasive fungal disease (IFD).

Conventional diagnostic techniques for fungal infection are invasive, slow to produce a result, and often lack sensitivity and species-level identification. 7 Direct evidence of fungi through evaluation of sterile specimens is the gold standard to prove IFD, 8 but requires potentially invasive procedures to obtain infected tissue.
Alternatively, noninvasive options that can suggest IFD include testing for galactomannan (GM), a cell wall component of Aspergillus species, and the glucose polymer 1,3--D-glucan (BG), a major cell wall component of most fungal species (with the exception of Mucorales and Cryptococcus). While these components may be released in blood and tissues in the course of invasive fungal infections, GM and BG do not detect all fungal pathogens and when studied in pediatric patients as a diagnostic tool, these have variable performance. Specifically, the GM test has a positive predictive value of 0-100% and negative predictive value of 70-100%, and the BG test has a positive predictive value of 17-49% and negative predictive value of 84-96%. 9 Radiographic findings are used to aid diagnosis of fungal pulmonary pathology in the setting of fever and neutropenia (FN), but are notably nonspecific, result in overdiagnosis, 10,11 and cannot accurately detect a specific fungal pathogen to best guide treatment. 12 PCR-based technologies are a promising approach to compliment current methods, but false-positive and -negative results continue to impede widespread applicability. 13 Targeted fungal sequencing (either 18S or ITS2) of biopsy tissue has reasonable performance with one institution showing sensitivity and specificity of 96.6% and 98.2% in patients with known diagnoses. 14 However, this approach requires an invasive procedure to obtain adequate specimen for testing.
Next-generation sequencing (NGS) technologies are being investigated for noninvasive diagnosis and monitoring of infectious diseases, including fungal pathogens. 15,16 Sequencing of cell-free DNA (cfDNA) in the bloodstream has previously demonstrated clinical utility in the detection of fetal abnormalities, 17 transplanted organ rejection, 18 and malignant tumors, 19 although results are disappointing for several major tumor types. 20 Similar approaches have been used to sequence circulating cfDNA, identify nonhuman sequences, and compare them with known genomic databases of bacterial, viral, and fungal pathogens. 18,21 While NGS technology is newer with a smaller number of samples studied, inhibiting comprehensive evaluation of its performance, this method has been effective for diagnosing bacterial infections in a real-time setting as well as noninvasively identifying molds in patients with invasive fungal infection. 16,22 In this pilot study (Clin-icalTrials.gov identifier: NCT03262584), we hypothesized that cfDNA NGS could accurately and noninvasively identify fungal pathogens in pediatric hematology, oncology, and HSCT patients at risk for new IFD.

Eligibility
Pediatric hematology, oncology, and HSCT patients treated at our institution and deemed at risk for IFD were eligible. Enrollment criteria included at least one of the following: (1) prolonged (≥96 h) FN despite broad-spectrum antibiotic therapy, (2) recrudescent FN, or (3) findings that triggered consideration of a new fungal infection (i.e., abnormal imaging or laboratory results). FN was defined as an absolute neutrophil count <500 cells/ L and axillary or oral temperature ≥38.5 • C or two temperatures at least 1 h apart in a 24-h period ≥38 • C. Recrudescent FN was defined as a second febrile episode after the first one resolved with antimicrobial treatment, on or after day 4 of empirical antibiotic therapy. Patients on prophylactic antifungal therapy were eligible for participation. Those patients without concern for fungal infection or those on >4 days of treatment-level antifungal therapy were excluded from the study given potential clearance of cfDNA from circulation on treatment-level therapy.

Trial design
The pilot study was an Institutional Review Board approved single center, prospective study conducted from May 2017 through May 2018 to determine if cfDNA NGS could identify fungal pathogens. Depending on patient age, eligible patients were contacted for consent (age ≥ 18 years), assent (age ≥ 12 years), and/or parental permission (age < 18 years). Patients could only participate in the study once. Of the 41 enrolled patients, initial blood samples were obtained from 40 patients with plans to collect a total of three samples (enrollment, follow-up 1, follow-up 2) during routine laboratory draws separated by approximately 1-month intervals. Follow-up samples were not collected from patients who were discharged from the institution or who passed away before study completion (n = 3), who did not have routine blood draws scheduled around time of follow-up (n = 4), or for whom followup was missed (n = 3). Clinical data, antimicrobial usage, and results from testing ordered by treating providers as part of standard care were obtained via electronic chart review; no study-specific testing was required for enrolled patients. Results from cfDNA NGS were not made available in real-time and did not direct decision making.

cfDNA NGS of infectious pathogens
Six milliliters of whole blood was collected in BD Vacutainer TM K 2 EDTA blood collection tubes (Becton Dickinson and Company, NJ) and plasma was separated within 24 h by centrifugation (1500 rpm for 10 min at room temperature). Two 1.5 cc aliquots were stored in sterile cryovials and frozen at -80 • C until transport to the Karius CLIA/CAP laboratory (Redwood City, CA) for processing. cfDNA was extracted from plasma, NGS libraries were prepared, and sequencing was performed on an Illumina NextSeq R 500. Sequencing reads identified as human were removed, and remaining sequences were aligned to a curated pathogen database. Any of over 1000 organisms in the Karius clinical reportable range found to be present above a predefined statistical threshold were reported as previously described. 16,23,24 Those interpreting the NGS results were blinded to clinical information. For detailed methods, see Supporting Information Appendix.

Patient characteristics
Baseline patient characteristics and eligibility criteria are shown in

Fungal classifications
Of the 40 patients who had enrollment samples drawn for cfDNA NGS, findings from chart review confirmed that six patients had proven IFD (

cfDNA NGS detects six fungal pathogens in patients with proven and probable IFD
cfDNA sequencing identified the same pathogen in four of the six proven IFD cases with fungi detected by invasive biopsy (skin, lung, pseudocyst fluid drainage) and blood culture (

cfDNA NGS negative at subsequent evaluations
Thirty-seven patients had follow-up samples drawn 1-month after their enrollment sample, and 30 patients had samples drawn at 2 months. Of the seven patients with fungal positivity on initial cfDNA NGS testing, including five pathogens defined in Table 2 along with P. jirovecii and C. glabrata, follow-up testing did not detect these organisms; all of these patients received at least a brief course of appropriate antifungal therapy. Overall, there were no fungi detected on follow-up testing for any patient.

cfDNA NGS detects infectious pathogens in patients with prolonged FN
Twenty-two patients were enrolled with prolonged FN despite broadspectrum antibiotic therapy and of these patients, two had proven IFD and six had possible IFD. While cfDNA NGS detected the same fungal pathogen in two patients with proven IFD and also detected C. glabrata in a patient otherwise not meeting any criteria for IFD, testing also revealed a variety of viral and bacterial pathogens in this population (   Ultimately, proving a diagnosis of IFD is difficult and routine practice promotes initiation of therapy based on concerning clinical or imaging findings, 40 especially in high-risk neutropenic patients with persistent or recurrent fevers despite broad-spectrum antibiotics. 41,42 In accordance with this practice, 21 of 28 (75%) of enrolled patients at risk for IFD due to prolonged or recrudescent FN were placed on >1 week of treatment dosing antifungal therapy by their primary providers. While other clinical or mycological findings may have supported this use, we determined prolonged antifungal therapy was used in 72% of patients who had no proven IFD and for whom cfDNA NGS testing was negative. If the negative predictive value of cfDNA NGS testing can be established, this could lead to decreased antifungal drug use since these drugs can have associated toxicities, lead to growth of resistant species, and increase medical costs. 43 For patients with high-risk FN, using a diagnostic test-guided preemptive strategy instead of an empirical approach to antifungal therapy may decrease overall antifungal use without increasing mortality. 44

CONFLICT OF INTERESTS
This is an investigator-initiated study with support provided by Karius,