Cell-free DNA as a Biomarker After Lung Transplantation

Chronic lung disease is the third leading cause of death in the United States. Lung transplantation is an effective short-term intervention for select patients with advanced lung disease. Nearly 4,000 adult lung transplant operations were performed in 2014 according to International Society for Heart and Lung Transplantation (ISHLT) registry data(1). AR, however, is more common among recipients of lung, as compared with other solid organ, transplants with approximately 35% of lung recipients experiencing at least one AR episode within first year of transplantation(1). Importantly, AR is the principal risk factor for chronic allograft dysfunction; the leading cause of late death in lung transplant recipients the primary obstacle to improved long-term lung transplant outcomes(2-5).

AR is distinguished upon observation of perivascular or peribronchiolar lymphocytic infiltrates in transbronchial lung biopsy specimens(6). Prior studies have demonstrated even low grade AR events (AR events of minimal or mild severity) increase the risk of chronic graft dysfunction, yet most of these episodes are asymptomatic(4, 5). As such, the majority of lung transplant centers perform surveillance lung biopsies at three-month intervals throughout the first posttransplant year and some continue annual surveillance biopsies over the lifetime of the allograft. Accordingly, lung recipients incur an enormous burden of invasive procedures and the inherent risks associated with this practice. There is a critical unmet need, therefore, to establish a less invasive biomarker of AR to minimize the number of biopsies after lung transplantation.

It has been proposed that damage to the lung allograft in AR disrupts the endothelial barrier, causing release of donor-derived cell-free DNA (dd-cfDNA) into the recipient circulation(7). This hypothesis is based on studies establishing dd-cfDNA as a useful biomarker of AR in heart transplant recipients(8). Specifically, a prospective study of heart recipients demonstrated dd-cfDNA facilitated the diagnosis of AR with a sensitivity and specificity comparable to that of endomyocardial biopsy(8). Similarly, a proof-of-concept study in a small number of lung transplant recipients demonstrated dd-cfDNA significantly increased in association with moderate or severe lung AR and also suggested dd-cfDNA increases in association with signs of chronic graft dysfunction(9). Gaps in this prior work lie in defining the utility of dd-cfDNA as a biomarker for mild/minimal AR events, which are far more common than moderate/severe AR and confer a similar increase in risk for chronic graft failure. Additionally, no prior work has sought to understand whether dd-cfDNA may correlate with treatment response in AR or whether dd-cfDNA can be detected in the lung allograft fluid. As such, it remains unknown whether lung fluid, obtained through BAL may be a more sensitive matrix for detecting clinical events in lung recipients.

WORK ON cfDNA TO DATE IN MULTICENTER LUNG TRANSPLANT COHORT (CTOT-20) CTOT-20 is an NIH funded, multicenter, observational cohort study enrolling lung transplant recipients at the time of transplantation with longitudinal follow up over approximately 5 years. The study enrolled over 800 lung recipients from 2015 to 2018 at five North American lung transplant centers. As part of CTOT-20, protocol mandated peripheral blood samples are collected at 1, 3, 6, 9, and 12 months posttransplant. The five centers enrolling in CTOT-20 all follow similar clinical care practices that call for surveillance BAL and transbronchial lung biopsies at intervals generally corresponding to these blood collection time points. Importantly, all biopsies performed on CTOT-20 subjects are assessed for the presence and severity of AR according to international guidelines(6) by an experienced lung transplant pathologist at the collecting center and these results are entered into the CTOT-20 eCRF. As such, we have accrued a unique biorepository of well-phenotyped paired plasma and BAL samples from lung transplant recipients.

As part of an NIH-funded CTOT ancillary study, we isolated total cfDNA from 320 paired plasma and BAL samples on 126 CTOT-20 participants. The majority of these samples were collected within the first \~6 months of transplant. A total of 60 patients in this cohort have at least one AR event represented, all of which are minimal or mild in severity. cfDNA was isolated using an automated nucleic acid purification system (Maxwell RSC, Promega). Up to 1ml of plasma or BAL was used. cfDNA yields were quantified using the Quantifluor (Promega) fluorescent dye system and isolated cfDNA is being stored at -20°C. TapeStation has been performed on the cfDNA isolated from approximately one-third of the BAL samples. This limited analysis suggests the majority of the BAL samples contain largely very small fragment DNA, without significant contamination by genomic DNA.

Prior studies have demonstrated even low grade AR events (AR events of minimal or mild severity) increase the risk of chronic graft dysfunction, yet most of these episodes are asymptomatic(4, 5, 10). As such, the majority of lung transplant centers perform surveillance lung biopsies at three-month intervals throughout the first posttransplant year and some continue annual surveillance biopsies over the lifetime of the allograft. Accordingly, lung recipients incur an enormous burden of invasive procedures and the inherent risks associated with this practice. There is a critical unmet need, therefore, to establish a less invasive biomarker of AR to minimize the number of biopsies after lung transplantation.

For the current collaboration, isolated cfDNA from the subjects/samples detailed herein (126 subjects, 320 BAL samples paired with 320 plasma samples, 640 samples in total) will be transferred to CareDx for determination of the dd-cfDNA fraction using the Allosure platform(11). Additionally, as it is completely plausible that the donor will not always be the lower contributor to the BAL cfDNA, we request CareDx perform genomic alignment of the paired BAL and plasma samples to inform the selection of the unknown contributor that most likely represents the donor fraction.

Results on the fraction of dd-cfDNA from the transferred samples will be returned to Duke for integration with the clinical data and statistical analysis supported by our CTOT-20 statistical team with sharing of results in a collaborative manner with the CareDx team. A detailed statistical analysis plan will be developed in collaboration with CareDx prior to data analysis. In general, our analytic approach for the descriptive and inferential analyses to address the primary hypothesis/outcome will be as follows:

For the descriptive analyses, biopsies will binned into the following groups based on time of measurement: 20 to 60 days / 61 to 140 days / 141 to 220 days / >220 days (roughly windowing 1, 3, and 6 months posttransplant). To describe the distribution of fraction of dd-cfDNA, the mean (standard deviation), the 5-number summary, and the deciles of levels will be computed. The distribution of fraction of dd-cfDNA levels will be described using a split-time boxplot by AR status and by transplant type across the biopsy time points. Pearson's correlation coefficient will be used to describe the correlation between paired blood and BAL fraction of dd-cfDNA.

For the inferential analyses, to test whether fraction of dd-cfDNA is elevated in transplant patients at the time of an AR event, fraction of dd-cfDNA levels will be modeled as a function of AR status and time from transplant using a linear mixed effect model with a random intercept and a random slope on time. Linear mixed modeling will be used to account for dependencies in the observations. The model will also adjust for potential confounders including transplant type (entered as a binary indicator of bilateral vs. single), time from transplant to biopsy (entered as a continuous covariate, centered and scaled). Additionally, concurrent infection will be entered into the model as a binary indicator. The model will be fit twice, once for plasma and once for BAL.