The Use of Molecular Techniques to Distinguish BK Nephropathy From Acute Rejection: Close but not Quite

Since the earliest report of polyoma (BK) nephropathy (BKN), clinicians have struggled to distinguish acute allograft rejection from active BK infections.1 While the histologic appearance of rejection can mimic BKN, the disparate treatment for each condition mandates an accurate diagnosis to enable the commencement of appropriate therapy. Until recently, the diagnosis of BKN remained relatively unchanged since the report by Purighalia,2 namely light microscopy findings of tubulitis, interstitial cellular infiltrates, antibody staining with anti-Simian virus 40 antibody and electron microscopic findings of 51-nm particles consistent with BK. In this issue of Transplantation, the report by Halloran et al,3 brings us a step closer to establishing an accurate diagnosis. For nearly 2 decades Halloran’s group has actively investigated the molecular and genetic characteristics of acute rejection and incorporated these features to complement standard histologic findings noted in kidney transplant biopsies that were diagnosed with acute rejection. In a series of large, prospective studies, they developed classifier algorithms to assess microarrays for T-cell and antibody-mediated rejection.4,5 With the use of their developed Molecular Microscope Diagnostic System (MMDx or “molecular microscope”) they were able to provide greater confidence to clinicians in the diagnosis of acute rejection compared to traditional histology. The present study extends these findings by attempting to use molecular identifiers to differentiate inflammation due to BKN infection versus cognate T-cell–mediated rejection (TCMR). Fifty kidney biopsy samples with confirmed BKN and 52 control samples without BKN were analyzed. Cognate TCMR was assessed using the MMDx. Major viral capsid 2 (VP2) mRNA was measured given the central role in BK transport, encapsidation, and propagation in BK infections. Using microarray expression and cross-validating predictions for BKN with machine learning, they attempted to develop a molecular classifier to estimate the probability of BKN. Overall, the authors were able to demonstrate VP2 was highly selective for BKN (area under the receiver operator characteristic curve = 0.94) and correlated this with acute injury, macrophage activation and the BKN classifier but not the TCMR classifier. The BKN probability classifier had a reasonable area under the receiver operator characteristic curve = 0.82. When they attempted to isolate specific human transcripts likely to be selective for BKN, they were unsuccessful. Of the 50 biopsy samples with BKN, a coexisting TCMR was found in 29 (58%) using MMDx; in those 29 cases, VP2 was positive in 28 (97%). Additionally, in 5 cases with serial follow-up biopsy samples, VP2 mRNA was decreased from the original value, but there was an increase in the TCMR score. The current Halloran study builds on previous work from their group using advanced techniques to examine molecular phenotypes of allograft infiltrates. Recently, several illuminating reviews elaborate on the “next generation of pathology.”6,7 The enhanced value of molecular and genetic analytics is evident, but the challenges of interpreting these newer diagnostic entities are also apparent. Several studies have recently been published with the goal of improving diagnostic accuracy in BK infections. Stervbo et al8 examined the use of T-cell receptor sequencing of allograft infiltrates in 1 patient to categorize the probability of BKN. In a larger study, Adam et al9 attempted to isolated gene sets seen in native kidney BKN, kidney allograft BKN, and TCMR. They determined that gene expression was a useful ancillary tool but not sufficient to distinguish concurrent BKN from acute rejection. In addition to the challenges noted, measurement of BK viremia is not standardized between labs and, in this study, was not utilized in the BKN classifier model. BK viral loads correlate with BKN,10 and, because of the patchy nature of this condition when examined histologically, it will be important in the future to study biopsy samples in patients with high viral loads but negative Simian virus 40 staining. Future larger studies will be needed to determine what is the appropriate threshold of VP2 (or other markers) when assessing for BKN. Lastly, and most importantly, prospective studies using composite scores of high-value markers for TCMR and BKN should include serial biopsies with outcomes. Where does this leave the clinician looking for clarity and reassurance when determining the diagnosis and treatment plan in a patient with graft dysfunction, BK viremia, and cellular infiltrates on the kidney biopsy? While we may be disappointed, it is not surprising, that the BKN classifier is not the answer. However, it is part of the answer and brings our field closer to where pathology is in other clinical arenas, such as oncology, where a combination of histologic findings with traditional light microscopy is supplemented with molecular markers and genetic transcripts. The challenge for pathologists and clinicians remains that BKN is heterogeneous, not binary. As this small series demonstrates, TCMR using MMDx is seen in more than half of cases of BKN. Even when isolated BKN exists, the accepted treatment is immunoreduction that when done too aggressively can result in acute rejection. As seen in the longitudinal biopsy cases, an expected reduction of VP2 but a rising TCMR score during treatment of BKN is not necessarily reassuring in a patient with deteriorating renal function following immunosuppression minimization. Ultimately, we need larger studies with the full spectrum of patients with BKN and acute rejection to discriminate these entities from each other. The future will likely include refinement of the BKN classifier, incorporating standardized BK PCR viral copies and gene profiles into the algorithm. While it is unlikely that any test/classifier/composite will give a binary diagnosis, transplant clinicians are well adept at utilizing multiple sources of data, examining our patient, and determining the most likely clinical diagnosis and management. The work by Halloran, and others, helps to move the field closer to certainty in BKN; we are just not there yet.