Intercenter Variability in the Application of Biopsy-based Transcriptomics in Kidney Transplantation

Abstract

Background.

Biopsy-based transcriptomics (BBT) was implemented in Central Europe in 2022 to improve kidney transplant diagnostics. Differences in diagnostic practices across transplant centers remain underexplored.

Methods.

This retrospective multicenter study analyzed 474 kidney graft biopsies from 10 transplant centers between August 2022 and May 2024, where, besides routine histology, BBT using the Molecular Microscope Diagnostic System (MMDx) was performed. Differences in BBT indications and discrepancies between histology assessment by Banff 2022 and MMDx sign-outs among transplant centers were evaluated.

Results.

Most centers used BBT in only 12%–31% of all performed biopsies, relying on histology alone for most diagnostic decisions. BBT indications varied across centers: 3 focused on histological no-rejection with clinical discrepancy (44%, 45%, and 70%), 2 on chronic or chronic-active antibody-mediated rejection (AMR; clinical discrepancy 44% and 48%), and 2 on borderline changes (clinical discrepancy 30% and 33%). BBT showed moderate agreement with histology (κ = 0.49), with similar discrepancy rates between high- and low-volume centers. Molecular AMR was found in 44% of probable AMR, 63% of active AMR, 63% of microvascular inflammation, C4d^– and donor-specific antibody (DSA)^–, 77% of chronic-active AMR, and 21% chronic AMR. Molecular T cell–mediated rejection (TCMR) was confirmed in 26% of histologically active TCMR, in 9% of chronic TCMR, and in 16% of borderline changes. In histologic no-rejection cases, molecular AMR was present in 10% of DSA^– and 34% of DSA⁺ biopsies.

Conclusions.

A moderate discrepancy between histology and MMDx sign-outs was found regardless of the center volume. BBT indications notably varied among centers. Standardized indications should be defined to improve the integration of molecular diagnostics into routine care.

INTRODUCTION

The Banff classification has remained the gold standard of histological assessment in kidney transplantation¹; however, biopsy interpretation can be subjective and dependent on the observer’s experience.² Biopsy-based transcriptomics (BBT) assessment has been suggested by Banff classification as early as in 2013³ to overcome heterogeneity and to improve transplant diagnostics. Despite this, its routine implementation has been slow and faced several obstacles. Among BBT platforms, the Molecular Microscope Diagnostic System (MMDx)⁴ has been implemented in the clinical routine across US centers with a centralized laboratory in Portland, whereas in Europe, clone laboratories in Prague and Vienna were recently opened in 2022 and 2024, respectively. Next, several centers in Central Europe have implemented BBT as a routine transplant diagnostic. The usage of BBT in clinical practice may vary across centers and needs to be further explored.

This retrospective multicenter cohort study aimed to evaluate the agreement between histology and BBT sign-outs, identify discrepancies, and assess differences in indications among centers during the initial period of BBT implementation into routine practice.

MATERIALS AND METHODS

Biopsies

Ten transplant centers from the Czech Republic and Slovakia have used BBT for transplant diagnostics since August 2022 (Table S1 and Figure S1A, SDC, https://links.lww.com/TXD/A790) using the central laboratory operating the MMDx platform in Prague (https://www.mmdx.eu/en/home). Aside from the sample for histology, another biopsy core or at least a 2–4 mm part of the biopsy was cut and placed into RNAlater for eventual BBT assessment. When clinicians indicated BBT, a biopsy sample was sent to the central laboratory at room temperature. A total of 474 MMDx assessments were performed till May 2024. Histological assessment was conducted in accordance with the latest Banff report.⁵ For discrepancy analysis between histology and MMDx, samples inadequate for histology diagnosis (n = 47) were excluded. Patients’ demographic data are shown in Table S2 (SDC, https://links.lww.com/TXD/A790). The Ethics Committee of the Institute for Clinical and Experimental Medicine approved the study protocol (approval No. 17210/24; A-24-11).

RNA Isolation and Microarray Analysis

RNA was isolated from biopsies, processed using GeneChip 3’ IVT PLUS Reagent Kit (Thermo Fisher Scientific, Carlsbad, CA), and labeled and fragmented cRNA was hybridized to the PrimeView Human Gene Expression Array (Thermo Fisher Scientific, Carlsbad, CA) as described previously.⁴ Microarrays were scanned using GeneChip Scanner 3000 7G (Thermo Fisher Scientific, Carlsbad, CA) and the resulting cell files analyzed using the MMDx algorithm with a reference set of 1208 kidney transplant biopsies.⁶

Statistics

Clinical characteristics were compared across groups using the Fisher exact test for categorical variables and the 2-sided Kruskal-Wallis test for continuous variables.

The discrepancy analysis between histology and MMDx was done in a subgroup of patients (n = 306) with proven antibody-mediated rejection (AMR), T cell–mediated rejection (TCMR), mixed rejection, and no-rejection by histology assessed by Banff 2022.⁷ Borderline changes, probable AMR, MVI, DSA^– and C4d^–, chronic TCMR, and chronic AMR were excluded from discrepancy analysis as these categories have no counterparts in MMDx sign-outs. Molecular rejection was defined as definite AMR/TCMR/mixed rejections. Subthreshold molecular TCMR was defined as mean TCMR classifier >0.1 and TCMR plus mixed archetype scores >0.2, and subthreshold molecular AMR was defined as mean AMR classifier >0.2 and sum of early AMR, fully developed AMR, and late AMR archetype scores >0.3.⁸ For the purpose of discrepancy analysis, molecular subthreshold rejections were counted as no rejections.

The Cohen’s weighted kappa value was calculated for discrepancies between histology and molecular results using library vcd (https://CRAN.R-project.org/package=vcd). Statistical analysis was performed using IBM SPSS version 24 (SPSS, Chicago, IL) and R Studio⁹ (version 4.2.2).

RESULTS

BBT Implementation in the Routine Practice

The clinical practice in the BBT implementation period widely differed among 10 centers (Figure S1A, SDC, https://links.lww.com/TXD/A790; Table 1). Kidney graft biopsies were indicated either for cause (n = 342; 72%) or per protocol (n = 132; 28%). In 7 transplant centers, biopsy specimens in RNA later were sent for BBT with the known histology, in 2 centers without the knowledge of histology, and in 1 center, both approaches were used.

TABLE 1.

Indications for BBT among centers

	TC1	TC2	TC3	TC4	TC5	TC6	TC7	TC8	TC9	TC10	P
BBT assessments, n	325	20	24	20	8	20	21	18	16	2
Biopsy POD	104 (3-11 026)	134 (13-3481)	102 (6-3160)	101 (8-2049)	292 (27-1568)	196 (6-4002)	1555 (50-5922)	141 (7-3131)	342 (15-5820)	93 (90-95)	0.001
DSA+	97 (31%)	11 (55%)	6 (25%)	12 (60%)	5 (63%)	6 (30%)	11 (52%)	8 (44%)	7 (54%)	0	0.012
Histology (Banff)
Probable AMR	11 (3.4%)	1 (5%)	1 (4%)	0	1 (12.5%)	1(5%)	0	1 (5.6%)	0	0
MVI, DSA–, C4d–	16 (4.9%)	1 (5%)	4 (17%)	0	2 (25%)	2 (10%)	1 (4.8%)	0	1 (6.3%)	0
Active AMR	18 (5.5%)	3 (15%)	4 (17%)	2 (10%)	3 (37.5%)	2 (10%)	3 (14.3%)	2 (11%)	1 (6.3%)	0
Chronic-active AMR	38 (11.7%)	3 (15%)	3 (13%)	2 (10%)	1 (12.5%)	2 (10%)	4 (19%)	0	7 (43.8%)	0
Chronic AMR	11 (3.4%)	0	1 (4.1%)	1 (5%)	0	0	6 (29%)	0	0	0
Borderline changes	18 (5.5%)	6 (30%)	8 (33%)	0	0	0	2 (9.5%)	2 (11%)	2 (12.5%)	0
TCMR	29 (9%)	3 (15%)	2 (8.3%)	1 (5%)	0	3 (15%)	0	4 (22%)	0	0
Chronic TCMR	21 (6.4%)	0	0	0	0	0	0	0	0	0
Mixed	11 (3.4%)	1 (5%)	0	0	0	0	0	1 (5.6%)	0	0
No rejection	100 (31%)	0	1 (4%)	14 (70%)	1 (12.5%)	9 (45%)	5 (23%)	8 (44%)	0	1 (50%)
Inadequate sample	39 (12%)	2	0	0	0	0	0	0	5 (31%)	1 (50%)
Othera	13 (4%)	0	0	0	0	1 (5%)	0	0	0	0

^aOther, pyelonephritis, BK nephropathy, IgA nephropathy recurrence, and necrosis.

AMR, antibody-mediated rejection; BBT, biopsy-based transcriptomics; DSA, donor-specific antibody; MVI, microvascular inflammation; POD, postoperative day; TCMR, T cell–mediated rejection.

In pediatric centers, BBT was performed in the majority of all biopsies, whereas in adult centers, BBT was used in discrepant cases only, according to clinicians’ discretion and the Banff recommendation for BBT assessment¹⁰ (Table 1). The number of BBT assessments reflected particular centers’ size (Table S1, SDC, https://links.lww.com/TXD/A790). Centers used BBT mostly <6 mo posttransplant (Figure S1B, SDC, https://links.lww.com/TXD/A790).

In total cohort, all categories of AMR (probable, DSA and C4d^– MVI, active, chronic active, and chronic) were the most prevalent histology when clinicians asked BBT (160/426; 37%) similarly to no-rejection in histology (139/427; 33%) and less frequently TCMR (61/427; 14%) and borderline changes (38/427; 9%; Table S2, SDC, https://links.lww.com/TXD/A790).

Indications of BBT widely differed among centers (Table 1). Interestingly, although some centers used BBT mostly when histology found no-rejection (TC4: 70%, TC6: 45%, and TC8: 44%), others indicated late biopsies with chronic or chronic-active AMR (TC7: 48%, TC9: 44%). Borderline changes were the reason for BBT assessment mostly in TC2 (30%) and TC3 (33%). BBT indications among centers are described in Table 1.

AMR

Among 158 biopsies with AMR in histology, molecular AMR was confirmed in 7 of 16 cases (44%) of probable AMR, in 17 of 27 cases (63%) of MVI, C4d^– DSA^–, in 24 of 38 cases (63%) of active AMR, in 47 of 60 cases (78%) of chronic-active AMR, and only in 4 of 19 cases (21%) of chronic AMR. Four biopsies were classified by MMDx as mixed rejections, and in 15 cases, the subthreshold molecular AMR was found. Conversely, 9 were reclassified as subthreshold TCMR (3 of them chronic AMR). Thirty-three of 160 cases (21%) of histologic AMR were molecularly no rejections (Figure 1A).

FIGURE 1.

Sankey plot of reclassification of Banff AMR (n = 158) (A) and TCMR (n = 101) (B) diagnoses by MMDx. The proportions of the histologic and the MMDx diagnoses are shown on the left and right sides of each diagram, respectively. Histologic AMR categories were defined as probable AMR, MVI, DSA^–, C4d^–, aAMR, caAMR, cAMR, and TCMR as BL changes, aTCMR, and cTCMR. Molecular diagnoses were defined as NR, pAMR, pTCMR, EAMR, FAMR, LAMR, mTCMR, and mixed. AMR, antibody-mediated rejection; aAMR, active AMR; aTCMR, active T cell–mediated rejection; BL, borderline; cAMR, chronic AMR; cTCMR, chronic TCMR; caAMR, chronic-active AMR; DSA, donor-specific antibody; EABMR, early AMR; FABMR, fully developed AMR; LABMR, late AMR; MMDx, Molecular Microscope Diagnostic System; mTCMR, molecular TCMR; MVI, microvascular inflammation; NR, no-rejection; pAMR, possible AMR; pTCMR, possible TCMR.

Borderline Changes

Of the 38 biopsies histologically classified as borderline changes, 6 (16%) were confirmed as molecular TCMR or subthreshold TCMR, 1 as mixed rejection, and 21 (55%) were molecularly classified as no-rejections. The remaining 10 of 38 biopsies (26%) were reclassified as AMR or subthreshold AMR (Figure 1B).

T Cell–mediated Rejection

Among the 42 biopsies categorized as TCMR in histology, 11 were confirmed as definite molecular TCMR, 2 as subthreshold TCMR, and 7 as mixed rejections. Conversely, 8 were reclassified as AMR (7 of them with v-lesion present), and 11 as no rejection (26%), 9 of which exhibited isolated arteritis lesions.

Interestingly, among 21 biopsies with chronic TCMR in histology, only 2 cases were confirmed as molecular TCMR, 3 cases as subthreshold TCMR, and 2 cases as mixed rejection, whereas 11 of 21 cases (52%) were molecularly classified as no rejection (Figure 1B).

No Rejection

Besides biopsies indicated for rejection in histology, those with no-rejection findings were assessed by BBT due to discrepancy with clinical findings. In DSA^– patients, 116 biopsies with no rejection in histology were examined (Figure 2A). The main histologic findings were isolated glomerulitis (n = 30), acute tubular necrosis (n = 19), IFTA (n = 19), or isolated tubulitis (n = 16). Molecular no-rejection was confirmed in 73% (85/116 cases), molecular AMR in 9%, molecular TCMR in 3%, and molecular mixed rejection in 2% of cases, respectively.

FIGURE 2.

Sankey plot of reclassification of no-rejection Banff categories in DSA^– patients (n = 116) (A) and DSA⁺ patients (n = 35) (B) by MMDx. The proportions of the histologic and the MMDx diagnoses are shown on the left and right sides of each diagram, respectively. Histologic NR categories were defined as ATN, isolated glomerulitis, isolated tubulitis, BKVN, IFTA, recurrent disease, pyelonephritis, and biopsies with no specific diagnoses (normal). Molecular diagnoses were defined as NR, pAMR, pTCMR, EAMR, FAMR, LAMR, mTCMR, and mixed. AMR, antibody-mediated rejection; ATN, acute tubular necrosis; BKVN, BK nephropathy; DSA, donor-specific antibody; EABMR, early AMR; FABMR, fully developed AMR; LABMR, late AMR; IFTA, interstitial fibrosis/tubular atrophy; MMDx, Molecular Microscope Diagnostic System; mTCMR, molecular T cell–mediated rejection; NR, no rejection; pAMR, possible AMR; pTCMR, possible TCMR.

In DSA⁺ patients, 35 biopsies with no rejection in histology were assessed (Figure 2B). Molecular no rejection was confirmed in 51% (18/35 cases), molecular AMR in 29%, and molecular TCMR in a single case as well as molecular mixed rejection. The majority of molecular AMR in DSA⁺ patients were acute tubular necrosis by histology during the early posttransplant period (5/10).

Inadequate Biopsy Specimens

In 47 cases, definitive histological diagnosis was not possible because specimens were considered less adequate or inadequate by the pathologist, and MMDx was thus performed to finalize the diagnosis. Among those biopsies, in 28 cases (60%), the MMDx revealed no-rejection, 4 (9%) AMR, 3 (9%) TCMR, 3 (9%) mixed rejections, and 8 (17%) subthreshold molecular rejection, respectively. These biopsies were not used for discrepancy analyses as histological diagnosis was not definitive.

Agreement Between Histology and MMDx Sign-outs

The unweighted kappa value for agreement between histology and MMDx in diagnosis of active (or chronic-active) AMR, active TCMR, mixed rejection, and no-rejection (after exclusion of subthreshold categories) was 0.49 (95% confidence interval [CI], 0.41-0.56) indicating moderate concordance (Table 2). There was no effect of center volume on the level of agreement between histology and MMDx sign-outs as similar kappa values were obtained; in high-volume TC1, the unweighted kappa value was 0.47 (95% CI, 0.38-0.57), whereas in all other centers, the unweighted kappa value was 0.50 (95% CI, 0.36-0.64).

TABLE 2.

Discrepancy analysis between histology and MMDx (N = 306)

MMDx	Banff histology				% Discrepancy (MMDx)
	AMR	TCMR	Mixed	No rejection
AMR	71	8	1	19	28
TCMR	3	11	4	5	52
Mixed	3	7	3	3	81
No rejection	21	16	5	126	25
% discrepancy (histology)	28	79	77	18
Unweighted kappa value	0.49 (95% CI, 0.41-0.57)

Banff histology: AMR, active or chronic-active AMR; TCMR; active TCMR; mixed, mixed rejection, no rejection, and all diagnoses without rejection. Borderline changes; MVI, DSA^– C4d^–, and probable AMR excluded.

MMDx: AMR (early-, fully developed, or late molecular AMR); TCMR, molecular TCMR; mixed, molecular mixed rejection, no rejection, molecular no rejection, or subthreshold rejection.

AMR, antibody-mediated rejection; BBT, biopsy-based transcriptomics; CI, confidence interval; DSA, donor-specific antibody; MMDx, Molecular Microscope Diagnostic System; MVI, microvascular inflammation; TCMR, T cell–mediated rejection.

BBT in Clinical Practice

To better understand the clinical application of BBT, a survey was sent out to 14 nephrologists from all participating centers (Table S3, SDC, https://links.lww.com/TXD/A790). The majority of responders (79%) considered the implementation of MMDx in the clinic as highly beneficial (5 points from a scale 0–5), 3 responders rating it as beneficial (4/5 points). In cases where MMDx results differed from histology, 36% of responders prioritized MMDx, 50% relied on any positive result (either histology or MMDx), and 1 responder relied solely on histology for treatment decisions.

The survey focused on histologically ambiguous diagnoses such as borderline changes, chronic AMR, isolated v-lesion, or isolated histological lesions (C4d, g, t) in DSA⁺ patients. In cases of borderline changes in histology, 71% of responders needed molecular TCMR for treatment decision, whereas the remaining 29% used BBT only in case biopsies for this purpose. Similar treatment patterns were reported for isolated v-lesions. None of the centers treated isolated histological lesions (lesions not fulfilling Banff 2022 rejection categories) in DSA⁺ cases when MMDx excluded molecular rejection.

DISCUSSION

In this multicenter observational cohort study, we describe wide differences among transplant centers in the use of molecular diagnostics since it became available in the region, underscoring the impact of institutional policies, biopsy protocols, and patient demographics on diagnostic practices. Some centers predominantly referred histological no-rejection cases with clinical discrepancies for BBT, illustrating a cautious approach in using molecular diagnostics to clarify ambiguous findings. In contrast, some centers prioritized late biopsies with chronic-active AMR, reflecting a focus on managing advanced rejection cases. Others indicated a preference of marginal categories, such as borderline changes. These variations highlight how biopsy indications are shaped by the clinical and logistical context of each center, including patient population characteristics, biopsy timing, and reimbursement policies. The volume of the transplant center affected the composition of patients’ cohort, with high-volume centers having wider patients’ variability, including highly sensitized and otherwise complicated cases.

All but 1 center used BBT in just a proportion of performed biopsies to improve rejection diagnostics, whereas in the majority of cases, histology alone was sufficient for patient management. The observed overall moderate concordance between histology and MMDx to diagnose active rejection (AMR, TCMR, mixed rejection) (kappa = 0.49) may reflect this practice and indicates the need for greater standardization. Notably, the discrepancies were consistent across centers, suggesting systemic rather than center-specific issues as histology was not centrally read and each center used its own pathology service.

The majority of discrepancies were found in boundary categories, such as borderline changes or probable AMR, which is consistent with previous studies.^11-13 However, discrepancies were also noted in chronic AMR and chronic TCMR. Nevertheless, discrepancies between histology and MMDx would have been much lower if the BBT had been indicated in all cases. In a cohort of 1679 biopsies from the INTERCOMEX study, the overall kappa value for the discrepancy between MMDx and histology was 0.41, which is comparable with our result¹⁴ and reflects the clinical use of BBT in problematic cases only.^15-17

There are several limitations of the BBT assessment. In this study, 5 of 7 cases with polyoma BKV nephropathy in histology were classified as molecular rejection, reflecting the known limitations of the method.¹⁸ In contrast, isolated arteritis in histology was often categorized as molecular no rejection, as it was described previously.^15,19 Yet, it remains unclear how to manage these cases, and the current discussion by the Banff community may answer these questions in the near future. The rejection treatment details and outcomes were not collected as this study was retrospective and observational, whereas centers differed even in practice when to indicate BBT.

The other important issue is the quality of biopsy samples. Inadequate biopsy specimens for complete histology diagnosis increase the probability of false-negative rejection findings. Interestingly, in our cohort, in 10 of 47 inadequate biopsies for histology, the molecular rejection was identified, which pointed out the clinical importance of BBT, because patients may refuse repeated biopsies. Despite molecular rejection being found in samples containing solely medulla, molecular scores for inflammation, cg score, and late-stage AMR can be overestimated.²⁰

To our knowledge, this is the first multicenter study that describes the clinical use of BBT. The findings emphasize the need to integrate molecular diagnostics more systematically into clinical workflows. Standardizing biopsy indications, training clinicians in the interpretation of molecular results together with histology and DSA results, may improve diagnostic consistency across centers. Additionally, addressing discrepancies in marginal categories, such as borderline changes, isolated v-lesion, and chronic inactive rejection, could further enhance the utility of molecular diagnostics in guiding clinical decision-making.

ACKNOWLEDGMENTS

The authors wish to express their gratitude to all clinical teams and pathologists for their valuable help in patient care.