Objectives: In the present study, we investigated donor-derived cell-free DNA dynamics and assessed the diagnostic efficacy of 2 tests: the sequencing of cytomegalovirus-derived cell-free DNA and the quantitative nucleic acid amplification test in cytomegalovirus infection following liver transplant.
Materials and Methods: We first examined 6 patients who were identified with active cytomegalovirus DNAemia by both quantitative nucleic acid amplification test and next-generation sequencing of cytomegalovirus-derived cell-free DNA and then performed a receiver operating characteristic analysis to evaluate the efficacy of cell-free DNA sequencing and establish a cutoff for this assay. Further validation of the next-generation sequencing method was also performed in 84 liver transplant recipients. The study protocol conformed to the ethical guidelines of the Declaration of Helsinki and the Declaration of Istanbul.
Results: In the first 6 patients, there was no significant correlation between the cytomegalovirus infection and donor-derived cell-free DNA. We determined that the levels of cytomegalovirus-derived cell-free DNA sequencing directly correlate with the results of the quantitative nucleic acid amplification test (area under the curve 0.982) and obtained a value of 0.015% as a cutoff for the cell-free DNA sequencing assay. In the validation cohort composed of 84 liver transplant recipients, next-generation sequencing of cell-free DNA revealed the occurrence of cytomegalovirus infection that remains otherwise undetected by the quantitative nucleic acid amplification test.
Conclusions: Cytomegalovirus infections that do not cause direct graft injury (cytomegalovirus-related hepatitis) did not result in elevations of donor-derived cell-free DNA. Next-generation sequencing of cytomegalovirus-derived cell-free DNA provides a potential tool for detection of cytomegalovirus infection that remains undetected by the quantitative nucleic acid amplification test.
Key words : Donor-derived cell-free DNA, Next-generation sequencing, Organ transplant, Quantitative nucleic acid amplification test
Cell-free DNA (cfDNA) has been exhaustively studied in recent years as a promising diagnostic, prognostic, and therapeutic monitoring tool in organ transplant.1-7 The analysis of donor-specific cfDNA provides an early, noninvasive yet direct measure to monitor allograft injury and rejection after transplant.1-3,7 There are many potential causes of allograft injury in liver transplant (LT), and standard clinical tests are often unable to reliably distinguish these common complications. Immunosuppression after LT is required to manage the risk of rejection but leaves patients vulnerable to infection. Infectious pathogens can be identified by simultaneous identification of nonhuman cfDNA sequences and comparisons of these with known genomic databases. Previous studies also demonstrated that sequenced donor cfDNA can be used to simultaneously test for rejection and infection after lung transplant.7 However, little is known about the diagnostic performance of infection monitoring with conventional pathogen tests and sequencing of cfDNA in LT.
Cytomegalovirus (CMV) is one of the most significant pathogens that negatively affects the outcome of solid-organ transplant recipients and confers higher risks of complications, graft loss, morbidity, and mortality.8,9 Cytomegalovirus primarily causes asymptomatic infection (positive for CMV DNA, but no clinical signs or symptoms) or CMV diseases (viral syndrome or tissue invasive disease) after LT. Overall, the rate of CMV diseases has ranged from 18% to 29% in LT in the absence of prevention strategy,10,11 and, even with the strategies of universal prophylaxis or preemptive antiviral therapy, 7% to 10% of LT patients will develop CMV diseases.10,12 The quantitative nucleic acid amplification test (QNAT) is the method recommended by The Transplantation Society guidelines to diagnose CMV infection, guide preemptive strategies, and monitor response to therapy.8 However, there is heterogeneity among QNAT results because of the lack of an international reference standard and variations in assay design.13,14 Surprisingly, even with the calibration of tests of the World Health Organization international standard, important differences still exist due to a variety of factors, including extraction method, amplification target, probe, nonstandardized quantification of secondary standards, and amplicon size.15-17
In the present work, we retrospectively analyzed the samples from 6 patients who were diagnosed with active CMV DNAemia. Next-generation sequencing (NGS) of plasma cfDNA from different timepoints was used to analyze CMV-derived cfDNA and donor cfDNA levels to investigate the dynamics of cfDNA and the relationship between CMV infection and graft damage. Further comparison between NGS results and QNAT results indicated that the analysis of CMV cfDNA by NGS was informative of CMV infection after LT (area under the curve = 0.982, sensitivity 100%, and specificity 88.10% at a Youden index of 0.881). We further investigated the diagnostic efficacy of 2 tests, the NGS of CMV-derived cfDNA and the QNAT, in 84 LT recipients. The results showed that NGS of CMV cfDNA revealed the occurrence of CMV infection that was undetected by QNAT.
Materials and Methods
The study protocol conformed to the ethical guidelines of the Declaration of Helsinki and the Declaration of Istanbul. This study was approved by the Ethics Committee of Second Xiangya Hospital of Central South University (No. 2019-050). Written informed consent was obtained from all participants. No prisoners were used in the study, and participants were neither paid nor coerced.
Subject enrollment and clinical sample collection
First, 6 patients who were diagnosed with active CMV DNAemia within the 3 months after LT at Second Xiangya Hospital (Hunan, China) were enrolled as the discovery cohort. Their plasma samples were collected and frozen at 0 days, 1 day, 3 days, 1 week, 2 weeks, 3 weeks, 1 month, 2 months, 3 months, as well as any timepoint when QNAT performed, and then thawed prior to cfDNA extraction and NGS. A total of 55 plasma samples were used in the discovery stage. A validation cohort was further recruited to evaluate the diagnostic efficacy of the NGS method for CMV-related infection. Plasma samples from this cohort were collected only when participants were clinically suspected to have CMV infection.
There were 95 LT patients consecutively enrolled in the first 3 months after receiving deceased donor LT between May 2017 and February 2018. Among these patients, 5 who received combined liver-kidney transplant (n = 2) or pediatric LT (≤14 years, n = 3) were excluded. Those who died within 3 months of transplant (n = 6) were also excluded. The remaining 84 patients included 65 males and 19 females, who had an overall mean age of 45.5 ± 11.90 years (range, 15-68 years). All cases of organ donation were performed according to the protocols for China Category I (organ donation after brain death).18 In the present study, a preemptive treatment approach was used to prevent CMV infection. Cytomegalovirus infection was monitored regularly for the first 3 months after transplant. All plasma CMV loads (viral loads) were measured with quantitative real-time polymerase chain reaction (PCR) assay and reported as international units per milliliter. Real-time PCR was performed on a quantitative PCR system (model Mx3000P, Agilent Technologies) with the CMV Real-Time PCR Kit (Daan Co. Ltd). Cytomegalovirus load equal to or above 200 IU/mL was defined as positive for viral replication and the threshold level for initiating preemptive therapy. Patients received standard-dose ganciclovir therapy (5.0 mg/kg every 12 hours for treatment dose, 5.0 mg/kg every 24 hours for maintenance/prevention dose) based on renal function. Cytomegalovirus antiviral therapy was continued for a minimum of 2 weeks, until clinical resolution of disease and eradication of CMV DNAemia below a threshold (<200 IU/mL).
All of the recipients received basiliximab as part of the induction protocol. Posttransplant immunosuppression included 500 mg intravenous methylprednisolone during the anhepatic phase. Tacrolimus was administered within 48 hours posttransplant at an initial dose of 0.05 mg/kg every 12 hours, with a subsequent dose adjustment to achieve whole blood trough levels. Mycophenolate mofetil was initiated on posttransplant day 3 at a dose of 500 to 750 mg twice per day for a total of 12 months and then discontinued.
DNA extraction and next-generation sequencing
For plasma samples, 10 mL blood was collected with Streck cfDNA BCT blood collection tubes (Streck) and centrifuged at 1600 g for 10 minutes at 4 °C to separate the plasma layer. A second stage of centrifuge was performed at 16 000 g for 10 minutes at 4 °C to remove any remaining cell debris. cfDNA was extracted with the QIAamp Circulating Nucleic Acid Extraction Kit (QIAGEN) from 3 mL plasma. A minimum of 50 ng DNA was used to prepare each NGS library per sample, with the NEBNext Ultra II DNA library preparation kit (NEB), according to the manufacturer’s instructions. Donor and recipient genomic DNA were extracted from blood samples with the phenol-chloroform method. A quantity of 1 μg of genomic DNA was fragmented and constructed to a NGS library for each sample. Libraries were then sequenced with an Illumina HiSeq X series sequencing system, and an average of 35 million reads per sample were generated with a PE150 sequencing strategy.
Donor DNA quantification
Sequencing data of genomic DNA from donor and recipient blood collected before surgery were used to identify sample-specific homology variations. In short, raw data were preprocessed to trim adapters and low-quality reads. Clean data were aligned to the human reference genome (hg38) with Bowtie 2 software. After deduplication, the primary alignment data were then processed with IndelRealigner and BaseRecalibrator of the GATK genome analysis toolkit pipeline. The recalibrated alignment data were used to identify specific variation sites that are homologous but with different alleles in paired donors and recipients.
The cfDNA sequencing data at different days posttransplant were then preprocessed as described above. Levels at the specific variation sites were dependent on the donor-specific allele. The donor DNA ratio was calculated by the levels of donor-specific alleles compared with the overall levels of the specific sites.
Cytomegalovirus load quantification
Raw sequencing data were first preprocessed as described above. Clean data were analyzed using our own Perl scripts. In short, clean data were mapped to hg38 by the Burrows-Wheeler Aligner BWA-MEM algorithm (see http://bio-bwa.sourceforge.net/). Unmapped reads were extracted and aligned to the CMV genome (NCBI GCF_000845245.1). The mapped reads were recorded, and the mapping ratio was calculated.
All statistical analyses and illustrations were performed with SPSS version 18.0 software. Receiver operating characteristic (ROC) analyses were performed using the ROCR package.
Levels of donor and cytomegalovirus-derived cell-free DNA during cytomegalovirus infection
To evaluate the relationship between donor and CMV-derived cfDNA in patient plasma during CMV infection, 6 patients who were identified with active CMV DNAemia within the first 3 to 6 months after LT were retrospectively reviewed. Their plasma samples collected and frozen at each timepoint of QNAT tests were thawed, followed by cfDNA extraction and NGS (Table 1). All 6 patients received liver allografts from deceased donors and were administered a preemptive treatment strategy; the symptoms of 5 patients with active CMV DNAemia were effectively controlled, and viral load was less than 200 IU/mL after 1 to 2 weeks of intravenous ganciclovir therapy. A 32-year-old discharged patient presented with acute onset of high fever, and follow-up at the outpatient clinic revealed that his liver enzyme concentrations were markedly elevated. The disease progressed rapidly, his aspartate aminotransferase concentration reached 2319 IU/mL, total bilirubin was 319.7 mmol/L, and prothrombin activity was < 20%. Liver biopsy revealed massive hepatic necrosis that stained positive for CMV protein. He was referred for CMV-related fulminant hepatitis but died while waiting for a new liver donor.
We examined donor cfDNA levels in these 6 LT recipients. The donor cfDNA percentages were highly elevated on the first days after transplant, most likely because of ischemia/reperfusion damage, which is consistent with other reports in the literature.19,20 However, the donor cfDNA percentage decreased within the first 1 to 3 weeks to a level of about 10% in 5 patients (patients A, B, C, D, and E) with no signs of liver injury or allograft rejection (Figure 1, A to E). The recipient with CMV-related fulminant hepatitis (patient F) died on day 46 posttransplant, and the level of donor cfDNA rose to more than 40% before death (Figure 1F, Table 1), which could reflect the dramatic liver damage caused by fulminant hepatitis.
We then quantified the CMV-derived cfDNA by mapping the sequencing reads to the CMV genome for each sample and investigated the relationship between CMV infection and donor cfDNA. Results showed that, except for patient F who was diagnosed with CMV-related hepatitis, there was no correlation between donor cfDNA and CMV-derived cfDNA. These results indicate that CMV infection itself does not correlate with elevated donor-derived cfDNA, partly because CMV infection may not directly result in graft injury (Figure 1).
Analyses of test performance
We next examined the concordance of viral load quantified by NGS and QNAT. During CMV DNAemia, the level of CMV-derived cfDNA calculated by NGS rose in correlation with QNAT result. After treatment, CMV cfDNA tended to decline; except for patient F, the CMV cfDNA for all of the other patients returned to baseline (Figure 2). To assess the efficiency of our NGS method to diagnose CMV infection, we performed a ROC analysis. All samples at each timepoint of the 6 patients were used as ROC input. Cytomegalovirus-positive samples were determined by QNAT result to be greater than 200 IU/mL. Under this criteria, 13 positive samples and 42 negative samples were inputted. The mapping ratio, which is calculated as the reads mapped to the CMV genome divided by the total nonhuman reads, was used as a diagnostic value. Receiver operating characteristic analysis showed that the area under curve of our NGS method was 0.982, with a 95% confidence interval between 0.954 and 1.009. When the mapping ratio cutoff was set to 0.015%, the ROC curve reached a Youden index of 0.881 with a detection sensitivity of 100% and specificity of 88.10% (Figure 3). Interestingly, those samples that were positive in the NGS method and negative in the QNAT method were all collected at timepoints near the timepoints for the QNAT-positive samples. This phenomenon indicates that the NGS method may be sufficiently sensitive to detect low-level CMV infection that is undetectable with the QNAT method. Based on the ROC analysis, we selected 0.015% as a cutoff value for the NGS method, consistent with the validation study (dashed blue line in Figure 2).
Further validation with expanded sample size
To further confirm the diagnostic performance of the NGS method, 84 patients who had undergone LT were monitored for CMV infection by both the NGS of cfDNA and QNAT. The patients with clinically suspected CMV infection were tested immediately (Table 2). According to the QNAT result (≥200 IU/mL) and the NGS method cutoff (≥0.015%), 67 patients tested negative for CMV infection by both the QNAT and NGS methods; these 67 patients also had no clinical evidence of CMV infection. Of the remaining 17 patients who tested positive for CMV infection by the NGS method, 15 also tested positive by the QNAT method. However, 2 patients (patient 2 and patient 12) tested negative by QNAT (Table 2). One of these patients (patient 2) exhibited classic symptoms (fever, malaise, leucopenia, or thrombocytopenia), and antiviral therapy was effective, which indicates clinical suspicion for active CMV DNAemia. The other patient (patient 12) had a clinical presentation of CMV colitis, and a hematoxylin and eosin-stained section of a colon biopsy demonstrated typical CMV inclusions. These results showed that NGS detection of CMV cfDNA is concordant with QNAT detection. Moreover, the NGS method is able to reveal the occurrence of CMV infection at a level that was not detected by QNAT.
Donor-derived cfDNA testing has been introduced into clinical use to monitor organ transplant recipients for rejection.4,21 Sequencing of donor cfDNA can be used to test for infection by simultaneous comparison of nonhuman cfDNA with known genomic databases of pathogens.2,7 This study is the first to compare the NGS method for testing CMV-derived cfDNA head-to-head with the gold standard QNAT method. The results of these 2 methods were coincident in 15 patients with CMV DNAemia and 67 patients who tested negative. However, 2 patients (patients 2 and 12) who were clinically diagnosed with active DNAemia tested positive for CMV infection by the CMV cfDNA sequencing method, whereas QNAT tests were negative (Table 2). In patient 12, CMV colitis was present and colon biopsy revealed typical CMV inclusions. Cytomegalovirus viral loads can be very low or negative in CMV colitis, and the fact that CMV DNA was found in the sequencing of cfDNA likely reflects a higher sensitivity for the NGS method versus the QNAT method. Patient 2 was merely clinically suspicious of active DNAemia, and no evidence was found to confirm the diagnosis. These results suggest that CMV DNAemia as measured by sequencing of cfDNA mirrors precisely what is found with routine quantitative PCR. The cfDNA sequencing assay described here provides a potential alternative approach for noninvasive diagnostic assay of CMV infection.
For comparison of QNAT, the assay presented herein has the potential to become an important noninvasive method for CMV surveillance after transplant, with consideration of the following items. (1) High incidence of infection in transplant recipients is difficult to distinguish from rejection after transplant, and sequencing of cfDNA provides a method to simultaneously monitor for rejection and infection in transplant recipients by sequencing cfDNA.7 (2) Unlike pathogen-specific testing, such as QNAT, the sequencing-based cfDNA test has the potential to simultaneously detect different sources of infections, including bacterial, viral, and fungal pathogens.2,22 (3) The sequencing method is quantitative, based on mapping to the known sequence of CMV, and it is precise and reproducible across laboratories and easy to standardize among laboratories.4,19,23 Thus, it is convenient to introduce a uniform strategy with the NGS method, including the definition of a cutoff value for prediction of CMV disease and initiation of preemptive therapy.
Donor-specific cfDNA seems to be an early, promising noninvasive diagnostic assay for the detection of acute allograft rejection and other graft-associated injuries after transplant.6,7 Infectious disease has been defined as a clinical manifestation of damage to the host that results from host-microbe interaction.24 In the 6 LT patients with active CMV DNAemia, except for patient F, who was diagnosed with CMV-related hepatitis (CMV infection causes direct allograft injury), there was no significant correlation between the level of donor-derived cfDNA and the clinical positive test for CMV. This is not consistent with reports by De Vlaminck and colleagues7 who described the significant elevation of cfDNA for lung transplant recipients who tested positive for CMV in bronchial lavage samples or serum. However, the incidence of CMV infection is higher and also associated with the transfer of a larger CMV load in lung transplant than in other solid-organ transplants, and CMV pneumonitis (direct injury of CMV infection) is quite common in lung transplant recipients in particular, as it develops in more than half of all lung transplant recipients over the first 3 years after transplant.25-27 Furthermore, the samples collected in the first 2 months were excluded in their study, but all of our samples came from the serum during the first 3 months posttransplant, and acute injury from surgery or ischemia/reperfusion had an effect on the results of donor cfDNA.19,20 Those studies suggested that the strength of correlations between donor-derived cfDNA and CMV infection may depend on the infection status, body sites, and severity of the disease.
In this study, CMV infections that do not cause direct graft injury (CMV-related hepatitis) did not result in elevations of donor-derived cfDNA. Next-generation sequencing of CMV cfDNA reveals the occurrence of CMV infection that remains undiagnosed by QNAT. The cfDNA sequencing assay described here provides a potential alternative approach for noninvasive diagnostic assay of CMV infection that will be particularly useful to monitor transplant recipients.
Volume : 19
Issue : 4
Pages : 331 - 338
DOI : 10.6002/ect.2020.0415
From the 1Department of Liver Transplantation, Second Xiangya Hospital, Central South University, Changsha, Hunan; and the 2Center for Medical Genetics, School of Life Science, Central South University, Changsha, Hunan, People’s Republic of China
Acknowledgements: This work was supported by grants from the Natural Science Foundation of Hunan Province (Grant No. 2018JJ2496), the National Natural Science Foundation of China (Grant No. 82070679), and the Hunan Province Science and Technology (Grant No. 2019GK5010). Other than described above, the authors have not received any funding or grants in support of the presented research or for the preparation of this work and have no further declarations of potential conflicts of interest.
Corresponding author: Zhongzhou Si, Department of Liver Transplantation, Second Xiangya Hospital, Central South University, 139 Renmin Road, Changsha, Hunan 410011, China
Phone: +86 137 87787805
Figure 1. Donor Cell-Free DNA Dynamics and Cytomegalovirus-Derived Cell-Free DNA
Table 1. Clinical Characteristics and Outcomes of 6 Patients With Active Cytomegalovirus DNAemia
Figure 2. Cytomegalovirus Load Results of Next-Generation Sequencing and Quantitative Nucleic Acid Amplification Test at Different Timepoints of 6 Liver Transplant Recipients
Figure 3. Receiver Operating Characteristic Curve Base on Mapping Ratio of Cytomegalovirus-Targeted Sequencing of Circulating Cell-Free DNA Samples
Table 2. Sequencing of Cell-Free DNA and Quantitative Nucleic Acid Amplification Test Analyses of Patients With Cytomegalovirus Infection in Liver Transplant.