SAN FRANCISCO (GenomeWeb) – Noninvasive shotgun sequencing can be used to detect organ transplant rejection without prior knowledge of the donor's genotype, according to researchers from Stanford University and Cornell University.
Building off previous methods that have demonstrated that shotgun sequencing of cell-free DNA shed in the bloodstream from the donor organ can identify rejection, the team demonstrated that the donor does not first have to be genotyped, which they said would make it more useful in a clinical setting.
The group described the method this month in PLOS Computational Biology and plans to work with the National Institutes of Health's Genome Research Alliance for Transplantation consortium to test their approach alongside other methods, including the gold standard of invasive biopsy.
Iwijn De Vlaminck, an assistant professor of biomedical engineering at Cornell and senior author of the study, said that previous work had identified that donor-derived cell-free DNA is a "good marker of transplant rejection" because increased levels of donor cfDNA indicates more damage to the graft and thus, rejection.
That previous work, in both heart and lung transplantation, relied on genotyping both the organ donor and the recipient in order to distinguish which cfDNA fragments were being shed from the donor organ, and which were naturally occurring from the recipient.
"Through our collaborations and discussions with clinicians, we learned that in many cases it's difficult for clinicians to get a hold of a tissue sample from the donor," De Vlaminck said. That led to the researchers discussing whether it would be possible to identify donor-derived cfDNA without knowing the donor's genotype.
Eilon Sharon, lead author of the study and postdoctoral research fellow at Stanford, said that the team came up with a computational approach that relies on knowledge of SNP allele frequencies in populations.
Similar to previous methods, the researchers first take a blood sample from the transplant recipient and perform shotgun sequencing of cfDNA and also genotype the recipient.
The team then built statistical models that took into account the frequencies of different genotypes in the population, the genotype of the recipient, sequencing error, and genotyping error as well as the actual sequence data from the cfDNA. The researchers relied on data gleaned from the 1,000 Genomes Project to build their statistical models.
"The idea was to find the most likely proportion of cfDNA that is donor derived given the recipient genotype and cfDNA sequences by iterating over different donor ancestral populations and accounting for sequencing and genotyping errors," Sharon said.
One key factor though, Sharon said, is that the model assumes that the organ donor and the recipient are not related. That is typically true in cases of heart or lung transplantation, but not in the case of bone marrow transplants or even kidney transplantation, he said.
"For those cases, we took another step and modeled the relationship between the donor and the recipient and calculated for that," Sharon said.
The statistical models help estimate the proportion of cfDNA fragments that are from the recipient and donor. The next step is then to determine what proportion of donor-derived cfDNA indicates transplant rejection.
De Vlaminck said that previous studies by the group and others have identified thresholds for heart and lung transplants but he said that studies in larger cohorts will be needed to further refine those thresholds.
The researchers compared their so-called one-genome model to the two-genome method, where both donor and recipient genotypes are known. Looking retrospectively at 382 blood samples taken at various time points from 51 lung transplant patients, the researchers found that the two models were highly correlated. Analyzing 435 samples from 59 heart transplant patients, they found that the two methods were still highly correlated, although not as close as for the lung transplants.
The heart cohort had lower levels of donor-derived cfDNA, so inferring the donor genotype was harder, the authors wrote in the study. However, they found that both methods were able to predict organ rejection. In the heart cohort, however, the one-genome method's accuracy was slightly reduced for predicting moderate rejection.
Next, they tested the model for individuals who had received bone marrow transplants from relatives. They analyzed 76 samples from eight bone marrow transplant recipients. Two donors were unrelated, but six were siblings of the recipients. They found that the statistical model was able to learn the relationship between the donor and recipient and that the one-genome method was comparable to the two-genome method.
John Sninsky, chief scientific officer at CareDx, which markets a targeted cfDNA NGS test for kidney transplant rejection, said that the method was "fascinating" and adds to the growing body of research demonstrating the "power of shotgun sequencing of cell-free DNA" for transplantation applications. CareDx purchased ImmuMetrix, which spun out from Stanford based on research from Stephen Quake's lab, where De Vlaminck when was a postdoc at the time, and others on sequencing cfDNA to diagnose transplantation. CareDx was not involved in this most recent PLOS Computational Biology study, however.
Ultimately, CareDx chose to commercialize a targeted sequencing version, AlloSure, and to focus first on kidney transplantations. Sninsky said that the targeted sequencing method enables host and donor cfDNA to be differentiated from each other without genotyping and without the statistical methods developed in this study. However, he added that the recent study highlights the potential of the shotgun sequencing approach to go beyond identifying organ rejection.
For instance, in the study, the Stanford and Cornell team sought to demonstrate that their method has the potential to distinguish between graft versus host disease (GVHD) and cancer relapse in bone marrow transplant recipients.
In bone marrow transplantations, GVHD can develop as a result of the newly transplanted immune system attacking the recipient. "In this case, rejection can occur to any tissue," De Vlaminck said. In addition, patients who receive bone marrow transplants often do so to treat cancer. If a patient is having problems post-transplantation, it can often be difficult to determine whether the patient's cancer has relapsed or whether the patient is suffering from GVHD.
The researchers noticed that in patients who developed GVHD, the levels of cfDNA dropped. The researchers then took it one step further, sequencing the blood cells, as well, and found that in patients who developed GVHD, recipient cfDNA and blood cell levels first dropped, but recipient cfDNA then increased. The researchers attributed this subsequent increase in recipient cfDNA to injured tissue and suggested that it could be used as a marker of GVHD.
Previous research by De Vlaminck when he was still at Stanford found that cfDNA sequencing can also identify changes to the recipient's virome. Specifically, he found that levels of anellovirus, a nonpathogenic virus that is abundant in humans, fluctuates depending on the patient's immunosuppressant dosage and whether the patient is experiencing rejection.
Shotgun sequencing of cfDNA could enable these broader types of applications of monitoring a patient's overall health post transplantation, Sninsky said.