Abstract
Background/Aim: Plasma Epstein–Barr virus (EBV) viral load measurement is prognostic in nasopharyngeal carcinoma (NPC) disease monitoring; however, a consensus measurement approach does not exist. This study characterized the clinical performance of metagenomic next-generation sequencing (mNGS), an unbiased sequencing-based assay distinct from polymerase chain reaction (PCR) or targeted sequencing approaches, in 73 peripheral blood specimens from 32 patients diagnosed with NPC. Patients and Methods: Samples were analyzed for plasma EBV viral load either by mNGS profiling or PCR-based assays (either LMP2 or BAMHI-W PCR) and compared to tumor presence by clinical assessment. Plasma mNGS-based EBV detection was quantified as reads per million (RPM). Results: Plasma mNGS displayed similar overall performance (100% sensitivity, 86% specificity, 92% accuracy) to BAMHI-W PCR (100% sensitivity, 86% specificity, 94% accuracy) and superior performance to the LMP2 PCR assay (36% sensitivity, 56% specificity, 45% accuracy). In a subset of 13 patients who underwent longitudinal analysis, plasma mNGS EBV RPM correlated with cancer recurrence (95%CI Pre-CRT=232.10±214; 95%CI Post-CRT=0.34±0.32; 95%CI difference=−231.70±214; *p=0.03, paired t-test), suggesting plasma mNGS exhibits potential for monitoring recurrence. Conclusion: Plasma mNGS is a distinct method for EBV titer measurement in NPC patients and more broadly, is a promising method for non-invasive monitoring of disease status for infection-associated malignancies.
Epstein–Barr virus (EBV) is the etiologic agent for endemic-type nasopharyngeal carcinoma (NPC) (1) and the presence of the disease is strongly associated with plasma EBV titers. This association has motivated the use of plasma EBV viral load as a biomarker for NPC disease status (2-6). Several assays have historically been used for EBV detection in cancer patients including immunofluorescence, EBV-encoded RNA (EBER)-1 in situ hybridization, semi quantitative polymerase chain reaction (PCR), quantitative competitive PCR, and real-time PCR (7). The current standard EBV DNA load assay is based on real-time PCR with discrepancies among different laboratories improved since the first international EBV WHO standard was established (8). However, challenges remain in assay standardization, inter-assay variability, and the accurate quantification of low titers. Indeed, there exists clinically significant variation among viral load assays based on commonly used targets for PCR amplification (polymerase-1 (Pol-1) gene, LMP2, and BamHI-W region). Consequently, results from different assays are not directly comparable despite active efforts at harmonization, hampering the widespread clinical application of plasma EBV as a liquid biomarker (9-11).
Despite their extensive success, alternate approaches to published PCR and sequencing based assays (12), which often require standardization and sample specific preparation, would complement existing approaches (13). We thus report our application of metagenomic next-generation sequencing (mNGS) (14, 15). a complementary approach conceptually and technically distinct from published PCR based and targeted sequencing panels, for measurement of plasma EBV levels in NPC patients (73 samples from n=32 patients). We found that plasma mNGS-based EBV levels are a reliable biomarker of disease status in patients with NPC, performing similarly when compared to two routinely used PCR-based assays. Moreover, the longitudinal trajectory of mNGS-based measurement of EBV levels correlated with clinically assessed disease status in pre-treatment and post-treatment settings across a subset of patients who were assayed at multiple timepoints (n=13). Our work suggests that plasma mNGS could provide another sensitive, versatile platform for measurement of plasma EBV as a serum biomarker for NPC disease status.
Patients and Methods
Patient selection and plasma sample processing. Patients who were diagnosed with pathologically confirmed NPC at a single institution and had available peripheral blood specimens between 2015-2021 were retrospectively reviewed. Baseline demographic, serum laboratory studies, and clinical outcome data were extracted from the medical record and institutional cancer registry. For patients meeting the above eligibility criteria, we collected plasma samples in the Department of Clinical Microbiology, University of California, San Francisco. To determine whether a patient demonstrated evidence of active disease, either pathologic confirmation or radiographic evaluation with either MRI or CT within four weeks of plasma EBV measurement was required to assess disease status. For imaging-based assessment, the formal read provided by a board certified radiologist was used to determine disease status. Similarly, for pathologic assessment, the formal report provided by a board-certified pathologist constituted the identification of active disease. For categorization of timepoint, any samples collected prior to beginning chemoradiation were considered “Pre-treatment”, any samples during radiotherapy were considered “Intra treatment”, the first post treatment sample was considered “Post treatment”, and all future samples were categorized as “Surveillance” if there was no radiographic/pathologic evidence of disease or “Recurrence” if radiographic/pathologic evidence of disease was noted. This study was reviewed and approved by UCSF’s Committee on Human Research and procedures were performed under an IRB-approved protocol #10-01116.
EBV LMP2 and BAMHI-W PCR assays. For EBV quantification by LMP2-PCR, a commercially available approach comprising plasma extraction in 400 μl plasma sample and 90 μl elution volume along with a 25 μl extraction volume (Qiagen, Hilden, Germany) was added into a master mix comprising primer and probe targeting the EBV LMP2 gene region (Roche, Basel, Switzerland) for each sample along with an internal EBV positive control. The BAMH1-W repeats PCR assay was conducted at Stanford University and only qualitatively reported as detected versus not detected.
Cell free DNA extraction, mNGS library preparation, and next-generation sequencing. For each plasma sample, 500 μl was centrifuged at 5,000 g for 10 min, then 400 μl supernatant was extracted using the EZ1 Virus Mini Kit (Qiagen) on a Qiagen BioRobot EZ1 Advanced XL (Qiagen) with a final 60 μl elution volume. A UCSF clinical laboratory internal control consisting of a mixture of 1.0 μl 99-191 bp oligonucleotides was added to each sample prior to extraction. The mNGS library preparation was carried out using a dual-index adaptor ligation protocol. In brief, a 25 μl extraction was prepared with 1.5 μl New England Biolabs (NEB, Ipswich, MA, USA) Next Ultra II End Prep Enzyme Mix and 3.5 μl NEB Next Ultra II End Prep Reaction Buffer at 18°C for 2 min followed by 66°C for 5 min. The resultant 30 μl End Prep Reaction Mixture was combined with an Adaptor Ligation master mix containing 1.3 μl NEBNext Adaptor for Illumina (0.6 μM), 15 μl NEBNext Ultra II Ligation Master Mix, and 0.5 μl NEBNext Ligation Enhancer. After incubation at 20°C for 15 min, 1.5 μl USER™ Enzyme was added and the mixture was incubated at 38°C for 15 min. Subsequently, 45 μl resuspended Agencourt AMPure® XP Beads (Beckman Coulter, Brea, CA, USA) were added. The ligation-bead reaction was incubated at 25°C for 10 min after which the plate was placed on the magnet for 3 min to separate beads. The supernatant was carefully discarded and the beads were washed twice with 200 μl 80% ethanol. After air drying for 5 min, the adaptor-ligated sample was eluted in 25 μl Tris-EDTA buffer. Twelve μl of adaptor-ligated DNA fragments and 8 μl of in-house barcode were PCR amplified using LC480 with NEB Q5 MasterMix (NEB) spiked in 0.6X SYBR™ Green I Nucleic Acid Gel Stain (Thermo Fisher Scientific, Waltham, MA, USA) under the following conditions: initial denaturation at 98°C for 45 s; 20 cycles of 98°C for 15 s, 63°C for 30 s, and 72°C for 90 s; final extension at 72°C for 60 s. The PCR product was cleaned using 32 μl Agencourt AMPure® XP Beads (Beckman Coulter) and the library was eluted in 17 μl EB buffer. The concentration of each library was determined using Qubit 3.0 Fluorometer (Life Technologies, San Francisco, CA, USA) and 2 μl DNA from each library were used to prepare a final pool and loaded on a HiSeq 2500 instrument (Illumina, San Diego, CA, USA) at 8 pM concentration.
Data processing and statistical analysis. The raw data were processed using the SURPI pipeline (16, 17). Briefly, the pipeline performs quality filtering, removal of reads aligning to human, alignment of reads to NCBI GenBank database, further filtering, and mapping of reads aligned to microbial genomes in order to estimate multiple metrics including c. (overage and normalized reads per million (RPM) for each detected microbial component. In this manner, EBV RPM values were extracted for all samples with detection of three viral genome regions required for EBV levels to be considered detectable and log2 transformed. To assess association between baseline clinical parameters and mNGS, the log2 transformed RPM values (outcome) were used to generate a separate univariate linear regression model for age, sex, or stage. In addition, non-linear quadratic associations were assessed between log2 RPM values and age, but not such associations were observed. Comparisons based on clinical status were carried out using either a two tailed Student’s t-test or a Mann–Whitney test for skewed data for two group comparisons or analysis of variance (ANOVA) for multiple group comparisons. For analysis with pre-treatment and post-treatment measurements from a single patient, a paired t-test was used. Two-sided p-values at p<0.05 were considered significant. To estimate diagnostic test performance metrics for mNGS, LMP2 PCR, or BAMH1W PCR, sensitivity, specificity, positive predictive value, negative predictive value, and accuracy were calculated as follows: The sensitivity is the proportion of true positives (positive by clinical assessment or imaging) that have a positive test result. The specificity is the proportion of true negative cases by clinical assessment or imaging that have a negative test result. The positive predictive value (PPV) is the proportion of positive test results that are true positive by clinical assessment or imaging. The negative predictive value (NPV) is the proportion of negative test results that are true negative by clinical assessment or imaging. The accuracy is the proportion of observations for which the test result and clinical assessment or imaging results agree. For patients with multiple pre-treatment measurements, test results were identical, and the first available measurement was used. GraphPad PRISM 9 (Boston, MA, USA), v9.5.1 and SAS 9.4 (Cary, NC, USA) were used for statistical analysis.
Results
To assess the utility of plasma mNGS in NPC, we retrospectively identified a cohort of 32 NPC patients who received chemoradiation at our institution (Table I). The median age of patients at diagnosis was 54 years with a slight male predominance (n=18, 56%), and 30 (94%) were non-metastatic. A total of 73 mNGS samples from all 32 NPC patients were analyzed by mNGS (Table II). Samples were distributed across multiple clinical timepoints [n=32 (44%) pre-treatment; n=15 (21%) post-treatment, defined as ≤4 weeks of completing chemoradiation; n=25 (34%) in surveillance, defined as >4 weeks after completing chemoradiation]. A total of n=53 samples (73%) were additionally tested using two PCR tests, either the LMP2 or BAMHI-W assay. The LMP2 assay was performed in 38 samples across 24 patients whereas the BAMHI-W assay was performed in 24 samples across 20 patients. Of 53 timepoints with a simultaneous clinical assessment of disease status (defined as radiographic or pathologic examination ≤4 weeks of serum assessment), n=35 (66%) had definitive evidence of disease at the time of EBV measurement.
We first analyzed mNGS data across all detected pathogen-associated reads, revealing a significant enrichment in viral read mapping associated with NPC disease status, with the majority mapping to the EBV genome (Figure 1). We next directly compared EBV reads per million (RPM) values as measured by mNGS to two commonly used PCR-based plasma EBV assays targeting either the LMP2 gene or the BAMH1-W element, demonstrating different results based on PCR method (Figure 2A). Analysis of plasma EBV titers’ association with clinical disease status demonstrated relatively poor performance of the LMP2 assay (Figure 2B), with an overall accuracy of 45%. In contrast, the BAMHI-W assay had an overall accuracy of 94% (Figure 2C). Plasma mNGS-based EBV measurement performed similarly with an overall accuracy of 92% (Figure 2D). To test for effects of baseline patient characteristics on mNGS titers, we evaluated EBV mNGS RPM based on patient age (Figure 3A), sex (Figure 3B), or AJCC stage (Figure 3C), and found no relationship of EBV RPM with age (p=0.06), sex (p=0.97), or stage (p=0.34).
Given the known changes in EBV levels during and after therapy, we evaluated longitudinal plasma mNGS EBV RPM values in NPC patients who had pre- and post-treatment measurements (n=13). These demonstrated decreased EBV RPM following completion of therapy (Pre-CRT Mean 232.10, SEM 98.21; Post-CRT Mean 0.34, SEM 0.14; Difference −231.70; 95%CI difference=−445.70 to −17.78; *p=0.03, paired t-test) (Figure 4A). More generally, stratification of mNGS-based EBV RPM values based on the time of assessment (pre-treatment, post-treatment, or surveillance) showed decreased EBV RPM values in the immediate post-treatment setting as compared to pre-treatment (Figure 4B). Evaluation of each individual patient with multiple timepoints supported a similar trajectory responding to treatment with the notable exception of a single case, which had detectable EBV RPM values in the absence of definitive disease (Figure 4C). Intriguingly, one patient with a false positive mNGS measurement (open diamond, Figure 4C) experienced persistent post-treatment headaches despite extensive workup and palliative treatments, suggestive of an underlying inflammatory disease process. Finally, in one patient who had multiple mNGS measurements following therapy completion, plasma mNGS EBV RPM values tracked with definitive evidence of disease recurrence on PET/CT (Figure 4D).
Discussion
Conventional NPC surveillance is based on physical examination, endoscopy, radiographic assessment, and/or invasive pathologic evaluation, all of which burden individual patients and the healthcare system. However, evolving technologies to measure minimal disease burden are increasingly sensitive, potentially enabling the use of non-invasive plasma biomarkers (3). Recent work strongly supports the utility of EBV plasma viral load measurements in NPC and similar approaches are rapidly evolving for human papillomavirus (HPV)-associated oropharyngeal cancer (4). Here, we compared the performance of plasma mNGS to that of two current PCR-based methods for measuring plasma EBV viral load in NPC patients. The mNGS approach appears robust, exhibiting a similar performance to the gold standard of PCR-based assays. Of note, our approach differs technically from currently utilized PCR or sequencing based approaches, and could be considered as a complementary approach potentially useful in low resource settings or for plasma samples not able to be processed by conventional workflows. More broadly, beyond its potential utility in this particular disease context consistent with previously described sequencing based approaches (5-7), the flexible, non-targeted nature of mNGS raises the possibility of a single diagnostic workflow to perform plasma surveillance of various infection-associated malignancies and illnesses, which are a major contributor to cancer burden worldwide (8, 9).
Study limitations. It is a single-institution study of clinical convenience samples, whose timing was not standardized. Thus, missing data at the level of simultaneous clinical assessment of disease status or the use of only a single assay at a given timepoint are obvious limitations. In addition, the typical limitations of retrospective study design are inherent to our results including potential for selection bias and the contribution of uncontrolled confounding factors. It should also be noted that this mNGS assay is institution specific. Additional prospective and/or multi-institutional assessment of the mNGS approach for EBV monitoring would be needed to establish generalizability.
Conclusion
Our data support mNGS as a novel, clinically meaningful approach for non-invasive plasma biomarker measurement in infection-associated malignancies such as EBV-associated NPC.
Footnotes
Authors’ Contributions
Conceptualization: CYC, SM, SSY. Data curation: HNV, KR, DR, SA, JWC, CYC, SM, SSY. Formal analysis: HNV, AAL, SF, SM, SSY. Methodology: HNV, AAL, SF, CYC, SM, SSY. Writing – original draft: HNV, SSY. Writing – review & editing: All Authors.
Conflicts of Interest
HNV, AAL, KR, DR, SA, JWC, SM: None. SF: US patent 11,515,006, Pathogen Detection using Next Generation Sequencing. CYC: research grants from Abbott Laboratories, BARDA contract 75A50122C00022, and CDC contract 75A50122C00022; inventor on US patent 11,380,421, “Pathogen Detection using Next Generation Sequencing”, Scientific Advisory Board member and stock options from Mammoth Biosciences, Inc., Poppy Health, Inc., and BiomeSense, Inc.; Co-Founder and Advisory Board member and stock options from Delve Bio, Inc. SSY: research grants from Merck, EMD Serono, BioMimetix, Nanobiotix; publishing royalties from UpToDate, Springer, and Elsevier.
Funding
None.
- Received June 22, 2024.
- Revision received October 7, 2024.
- Accepted October 15, 2024.
- Copyright © 2024 International Institute of Anticancer Research (Dr. George J. Delinasios), All rights reserved.
This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY-NC-ND) 4.0 international license (https://creativecommons.org/licenses/by-nc-nd/4.0).