The “Borderline” Renal Allograft Biopsy in the Era of Molecular Diagnostics: A Sampling Conundrum?


Parmjeet Randhawa,


The author comments on the potential causes and clinical significance of discrepancies between conventional histologic and modern molecular approaches for evaluating renal allograft biopsies. See article by de Freitas et al on page 191.

A diagnostic category termed “borderline (BL) changes suspicious for acute rejection” was coined in the 1991 Banff Conference of Allograft Pathology to refer to biopsies that do not fulfill the histologic criteria needed to diagnose acute T-cell mediated rejection (TCMR; Ref. 1). The suggested management approach was to rule out other causes of graft dysfunction and to treat for rejection if spontaneous resolution did not occur. In this issue of the journal, de Freitas et al. present a study which suggests that molecular diagnosis can potentially eliminate the BL category by reclassifying all cases into discrete TCMR and non-TCMR categories (2).

The authors’ contention is on the basis of a study of 40 BL biopsies of which 13 (33%) were reassigned to the TCMR category on the basis of a molecular risk score >0 and a T-cell burden >2.32. Conversely, 27 (67%) BL biopsies were reassigned to the nonrejection category. For this latter group of biopsies, the authors imply that the molecular phenotype is the more correct one. However, as molecular phenotyping was done on a different tissue core than the one used for histologic studies, absence of the threshold T-cell signal necessary to diagnose molecular TCMR could simply reflect the well recognized focal nature of allograft inflammation and tissue injury. Histologic studies that compare the pathology in two simultaneously taken cores of tissue find significant core to variation. The most extreme example of this phenomenon is a study where cytopathic effect because of polyomavirus was seen in only one of two cores in 37% of biopsies examined (3).

The “histologic-BL” but “molecular no rejection” category identified by the authors requires further clinical validation. In the current series of cases, serum creatinine improved on follow-up regardless of diagnosis and steroid administration. This reflects the empirical nature of managing graft dysfunction in the clinic. Patients with or without biopsy proven rejection can receive any combination of volume replacement, adjustment of calcineurin inhibitor dose and steroid therapy, at times even before the official biopsy report has become available. Designing a validation study that keeps track of and statistically controls for these multiple parameters will be challenging.

Although the authors have chosen only to highlight histologic BL biopsies that had molecular features of TCMR, perusal of their data also identifies cases where conventional histology led to a diagnosis of TCMR that could not be picked up by molecular testing. Indeed, in 10 of 35 (28.5%) cases with histologic TCMR the molecular probability of rejection was judged to be less than 0.5 and 9 of 35 (25.7%) such cases did not show high T-cell burden. These discrepancies very likely reflect sampling issues that are inherent in renal allograft biopsy technique. Histologic parameters such as percentage inflamed cortex as well as molecular features such as total T-cell mRNA signal are a function of the size and location of the sample obtained for analysis (Figure 1).

Figure 1.

Sampling problems inherent in the renal allograft biopsy. (A) Is a diagrammatic representation of an allograft kidney with TCMR wherein the inflammation scores (i0-i3) and tubulitis scores (t0-t3) show a gradation from peripheral to central parenchyma. If a biopsy sample is taken for light microscopy (LM, thin arrows) and a second parallel sample is taken for molecular studies (thick arrows), the degree of inflammation captured may be worse in LM (sampling toward the top), concordant in both samples (sampling in the center) or better in LM (sampling toward the bottom). (B) Conceptualizes sampling problems that will result if distribution of TCMR in the kidney is assumed to be focal (dark shade). Depending on the actual site of sampling, inflammation and tubulitis will affect 30–100% of a 2-cm long biopsy core taken from different regions. The percentages affected will change further for any given site if a smaller 0.5-cm core located in the middle of the longer core is sampled instead. For the topmost 2.0 and 0.5 cm biopsy pair, the captured T-cell burden (length of the shaded portion of the core) will be independent of the percentage inflamed area visualized by light microscopy. However, for the lowest site of sampling, the T-cell burden assessed by raw gene expression signal value would be greater for the longer core than the shorter core, unless results are normalized to a housekeeping gene that has been shown not to be altered during acute rejection. Conversely, if such normalization is done the smaller of the two cores in the topmost biopsy pair will appear to have a higher T-cell burden per unit length of tissue. Thus, although molecular studies sample much more tissue than a conventional 3–5 μm histologic section, they cannot compensate for sampling heterogeneity that results from the focal nature of rejection-associated infiltrates. This is to be expected because a typical biopsy measuring 10 × 2 × 2 mm has an approximate volume of 40 cubic mm, which constitutes only about 0.03% of the total parenchymal volume of a typical adult kidney assumed to measure 12 × 5 × 2 cm.

A particularly instructive point made in this study is that Banff-BL biopsies with TCMR-like molecular features had similar i-total scores, lower i-Banff scores and higher interstitial fibrosis/tubular atrophy scores, compared to biopsies with an unequivocal histologic diagnosis of TCMR. This observation calls into question two important cornerstones of the Banff Schema as currently implemented: (a) areas of atrophy and scarring should be ignored for the purposes of obtaining i- and t-scores and (b) a minimum threshold of t2 tubulitis be required to make a definitive diagnosis of TCMR. The authors’ data make a strong case for these guidelines to be revised. Specifically, it is shown that in biopsies with scarring, inflammation affecting greater than 27% of the biopsy area corresponds nicely with the molecular diagnosis of TCMR in about 85% of cases. For biopsies with i < 27%, a molecular signature of TCMR was found in approximately one half of biopsies wherein the extent of tubulitis (irrespective of its grade) exceeded 3% of the biopsy area.

Revisiting the diagnostic criteria for TCMR in biopsies with interstitial fibrosis/tubular atrophy should be an important future priority of the Banff Working Group. It is worth recalling that a 5% inflammation threshold and tubular injury per se (rather than tubulitis grade) were the key features of a biopsy interpretation schema used by the collaborative clinical trials in transplantation group (4). Quantitation of intraepithelial lymphocytes is not a criterion that is used either for the diagnosis or grading of TCMR in other solid organs, including the liver, pancreas and small intestine. Therefore, the prime importance given to severity of tubulitis in the grading of renal allograft rejection should be reconsidered, particularly in the context of late transplant biopsies.


The author of this manuscript has no conflicts of interest to disclose as described by the American Journal of Transplantation.