All KTX recipients (>18 years) at the outpatient clinic of the Department of Nephrology and Dialysis at the Medical University of Vienna between 11/2022 and 02/2024 were eligible for inclusion if they met one of the following criteria: Detection of BKPyV and/or JC polyomavirus (JCPyV) viremia on routine PCR; identification of decoy cells in routine urine sediment; a urinary viral load of BKPyV and/or JCPyV >10⁸ U/mL; biopsy-proven BKPyVAN (BP-BKPyVAN), which was performed in line with current consensus recommendations in the context of allograft dysfunction [11]. Patients were included at the time of clinically relevant BKPyV reactivation prompting diagnostic evaluation, rather than at a predefined virological stage, resulting in variable timing relative to transplantation and viral kinetics. In addition, a control group of clinically stable KTX patients without evidence of relevant BKPyV/JCPyV replication was included. Controls had no detectable DNAemia and no clinically relevant DNAuria according to current consensus definitions (urinary DNA <10⁷ copies/mL). The study flowchart is presented in Figure 1.

Urinary and plasma BKPyV and JCPyV-DNA were quantified as part of routine post-transplant surveillance at our center in all kidney transplant recipients included in the study, following international screening recommendations, acknowledging that PyVAN may rarely be caused by JCPyV (for details, see Supplementary Materials) [25, 26].

BKPyVAN was categorized according to the most recent international consensus framework into four grades: possible, probable, presumptive, and definite BKPyVAN [13] (see details for each diagnostic tier in the Supplementary Materials).

Urine cell pellets were ethanol-fixed and immunolabeled for BKPyV capsid protein VP1, with DAPI nuclear counterstain, and acquired on a FACSCanto flow cytometer [16]. Flow-cytometry data were processed using custom Python scripts in JupyterLab. Samples with marked granulocyturia were pre-specified as technical exclusions. The detailed protocol is provided in the Supplementary Materials.

Urine sediment cells were maintained in culture medium as described above (RPMI 1640 supplemented with 2% FCS and ampicillin-streptomycin). A 60 μL aliquot was applied to the funnel of a cytocentrifuge and centrifuged at 1,200 rpm for 3 min. The resulting cytopreparation was air-dried for 1 h, followed by fixation in acetone for immunostaining or in the fixation solution of the Hemacolor^Ⓡ (Merck, Merck KGaA, Darmstadt, Germany; Millipore, SIG 1116740001) three-component kit for 4 min (see Supplementary Materials).

Cytopreparations were acetone-fixed and immunostained for BKPyV VP1 using either a mouse monoclonal primary (Invitrogen 4942, 1:60) or a rabbit anti-VP1 primary (1:1,000) on separate slides. After incubation (overnight at 4 °C or 2 h at room temperature) and PBS washes, Alexa Fluor 594 goat anti-mouse or Alexa Fluor 488 goat anti-rabbit secondary antibodies (1:700) were applied with DAPI nuclear counterstain. Slides were mounted in Vectashield and imaged on a Zeiss inverted confocal microscope; images were post-processed in Adobe Photoshop. Full step-by-step details are provided in the Supplementary Materials.

One cytopreparation was sequentially stained with eosin, followed by hematoxylin and a washing buffer, using the Hemacolor^Ⓡ staining kit (Merck, Merck KGaA, Darmstadt, Germany; Millipore, SIG 1116740001). Cell enumeration was performed via light microscopy (see Supplementary Materials).

Continuous variables were assessed for normality with the Shapiro-Wilk test and are presented as means ± standard deviations (SD) when normally distributed or as medians with interquartile ranges (IQR) for skewed data. Group comparisons were performed using the student’s t-test or the Mann-Whitney U test. Kruskal-Wallis’s test was used to compare the differences in BKPyV-DNAemia and urinary-VP1-FC counts in more than two groups. Complete statistical analysis is described in Supplementary Materials.

Between November 1st^, 2022 and February 29th^, 2024, 30 KTX recipients meeting our inclusion criteria for the study group and 21 KTX for the control group were analyzed. A flowchart of patient inclusion and exclusion is depicted in Figure 1. Table 1 provides all relevant patient dispositions and transplant-specific baseline characteristics. Median age was 59 years (IQR 49.8–64.0) and 19 (63%) recipients were male. Donor-specific antibodies were detected in one (3%) patient. The median HLA A/B/DR mismatch was three (IQR 2–4). As induction therapy the majority (97%) received IL-2 antibody basiliximab (20 mg on days 0 + 4). Maintenance immunosuppression consisted of tacrolimus/MMF/steroids in all patients (Table 1). The first flow‐cytometry assay was performed at a median of 11.5 months after KTX (IQR 2.7–18.1). At baseline, BKPyV-DNA was detectable at high levels in urine (median 3.3 × 10⁶ c/mL, IQR 2.5 × 10¹–4.0 × 10⁹) and well detectable in plasma (median 4.6 × 10³ c/mL, IQR 2.5 × 10²–8.8 × 10³). Tacrolimus trough levels at the time of the first urinary VP1-FC sample (baseline) were at a median of 7.8 ng/mL (IQR 6.9–9.3), which was followed by a constant reduction due to significant BKPyV infection. In 29 (96%) patients the MMF dose was either reduced or MMF was paused. Serum creatinine mirrored this, with a median of 1.86 mg/dL (IQR 1.09–4.82) at baseline and a subsequent increase from months one to three, which was afterwards followed by a stabilization phase (Table 1; Supplementary Table S1). Renal allograft biopsy was performed in 21 of 30 patients (70%) with 10 being classified as biopsy-negative and 11 as biopsy-proven BKPyVAN. With respect to the timing of the biopsy, eight (38%) were performed at a median of 1 month (IQR 1.0–2.0) before the first urinary VP1-FC assessment - of which none showed signs of BKPyVAN - and 13 biopsies (62%) were performed at a median of 1 month after urinary VP1-FC (IQR 0.0–4.0, Supplementary Figure S1). Protocol biopsies accounted for 40% (4/10) of the biopsy-negative group and 18% (2/11) of the biopsy-proven group (p = 0.36). Three of the eight biopsies performed prior to VP1 sampling were protocol biopsies.

Our control cohort comprises 21 KTX recipients screened during the study, of whom 62% were male, with a median age of 63 years (IQR 50.0–67.5). All had negative BKPyV-DNAemia and no clinically relevant BKPyV-DNAuria (DNA-uria centered at zero; median 0 copies; IQR 0–145) and low JCPyV-DNAuria (median 0 copies; IQR 0–2,850). Correspondingly, FC measurements were low and clustered around baseline, with a median value of 0.0 (IQR 0.0–3.2). In the control cohort 9 (43%) biopsies were performed, 7 (33%) after the VP1-FC sampling. All were negative for BKPyVAN.

Urinary-VP1-FC using the rabbit polyclonal anti-VP1 antibody detected a median of 13% (14%–29%, n = 30) VP1-positive urinary epithelial cells. When the same cytospin was examined by indirect immunofluorescence, the rabbit-derived immunoreagent produced a median signal of 10% positive cells (0%–84%, n = 30). In contrast, the subset stained with the mouse monoclonal antibody averaged 6% (2%–40%, n = 19). Classical hematoxylin-eosin morphology identified decoy-like nuclei in 3.5% (0%–40%, n = 30) of cells, and the pathologist’s semi-quantitative decoy cell scoring showed a median of 2% (0%–25%, n = 21) decoy cells. In the BK-negative control group, all baseline cross-assay measurements were centered at background levels, with median values of 0.0 for VP1-FC, indirect immunofluorescence, and hematoxylin–eosin assessment. Supplementary Figure S2 shows the representative confocal immunofluorescence of urine sediment from a 58-year-old male KTX recipient with BKPyVAN, with numerous VP1-positive cells at low magnification. The detailed fluorescence intensity measurements of five representative transplant recipients are provided in Supplementary Figure S3, presenting the technical performance of the urinary VP1 assay.

Among 30 recipients stratified by baseline plasma BKPyV-DNAemia (Figure 1), urinary VP1-positive cell percentages differed significantly across groups (p = 0.04). In Dunn’s post-hoc comparisons, recipients with presumptive BKPyVAN showed a greater VP1 burden than those with possible BKPyVAN (adjusted p = 0.03). When BK-negative controls were included in the same model, controls showed consistently lower VP1 levels than all BKPyVAN categories (Figure 2A). Among recipients stratified by diagnostic certainty (possible BKPyVAN with low-level BKPyV-DNAemia, probable BKPyVAN, and biopsy-proven BKPyVAN), urinary VP1-positive cell percentages increased progressively across tiers, with a significant overall group difference (Figure 2B; p < 0.001). Plasma BKPyV-DNAemia likewise varied significantly by diagnostic tier (Figure 2C; p < 0.001). Stratification by biopsy status demonstrated significant differences in urinary VP1 burden (Figure 2D; p = 0.0009) and plasma BKPyV-DNAemia (Figure 2E; p < 0.001). Finally, patients with isolated BKPyV-DNAuria had significantly lower urinary VP1 positivity than those with concurrent BKPyV-DNAemia (p = 0.02).

Across the five flow-cytometric assessments, the distribution of VP1-positive cells became progressively narrower and shifted toward lower values (Figure 3). Collection timepoints (TPs) for the FC samples are described in Figure 3. At baseline (TP1), the median VP1 burden was 13% (IQR 4%–29%). At TP2 the median fell to 11% (IQR 4%–31%), then to 6% (IQR 1%–36%) at TP3, and to 0.6% (IQR 0%–4%) at TP4. At TP5 the median had declined to 0% (IQR 0%–0.4%); p < 0.001. Collectively, these data indicate that the bulk of VP1 clearance occurred between months two and four post-urinary-VP1-FC sampling (i.e., between TP3 and TP4), after which the viral signal plateaued to nearly background levels in almost all patients (Figure 3A). Parallel to the flow-cytometric quantification of VP1-positive cells, at TP1, the median viral load was 4.0 × 10³ copies/mL (IQR 4.0 × 10²–9.0 × 10³), which declined to 2.0 × 10³ copies/mL (IQR 1.5 × 10²–2.8 × 10⁴) at TP2 and to 1.3 × 10³ copies/mL (IQR 4.0 × 10²–2.9 × 10³) at TP3. By TP4 the median had fallen further to 5.0 × 10² copies/mL (IQR 1.3 × 10²–1.0 × 10³) and at TP5 to 4.0 × 10² copies/mL (IQR 2.0 × 10²–1.0 × 10³), despite complete loss of VP1 signal by FC at TP5. Figure 3B, shows a rapid initial drop in DNAemia between TP1 and TP2, followed by a gradual plateau towards low-level DNAemia thereafter. Across visits, urinary VP1% showed a positive correlation with BKPyV-DNAemia. At TP1, the correlation was moderate (rho = 0.59, p = 0.001), but it was strong at TP2 (rho = 0.75, p = 0.01). At later visits, correlations were weaker and not statistically significant.

Receiver-operating-characteristic analysis for urinary VP1-FC, using biopsy-proven BKPyVAN as the endpoint of interest and restricted to 20 recipients biopsied after study inclusion (including 7 available biopsies from the control group), demonstrated excellent diagnostic performance when using baseline (TP1) values. The AUC was 0.98 (95% CI 0.93–1.00, p = 0.0003; Figure 4A). A VP1 threshold of 7.4% achieved 100% sensitivity and 88.9% specificity. Increasing the cutoff to 11.7% resulted in a sensitivity of 90.9% and specificity of 88.9%, whereas a stricter rule-in threshold of 17.5% yielded sensitivity of 81.8% with 100% specificity, thereby favoring confirmation over detection. In LOOCV, urinary VP1-FC retained good discriminatory performance (AUC 0.89, 95% CI 0.68–1.00; p = 0.0034) despite the small cohort (Figure 4B). However, when separating groups based on possible, probable, and presumptive BKPyVAN using viremia levels at TP1, AUC for the prediction of BKPyVAN by urinary VP1-FC applying a cutoff of 17.5% was 0.74 for possible, 0.52 for probable, and 0.80 for presumptive BKPyVAN. Corresponding values for sensitivity were 47%, 33%, and 83%, and for specificity were 73%, 56%, and 71%. The optimal VP1‐burden thresholds maximizing Youden’s index were >10% for possible BKPyVAN (sensitivity 79%, specificity 73%), >4% for probable BKPyVAN (sensitivity 92%, specificity 33%), and >40% for presumptive BKPyVAN (sensitivity 67%, specificity 96%). The secondary analysis using the combined logistic model of urinary VP1 and BKPyV-DNAemia demonstrated the strongest predictive performance for biopsy-proven BKPyVAN (n = 13, AUC = 0.97). Predicted probability surfaces revealed that probabilities exceeded 0.95 already at approximately 18%–20% VP1 positive cells, even at low viral loads, and approached 0.99 beyond 25% VP1 across any BKPyV-DNAemia (Supplementary Figure S4).

A total of 68 urine samples from 30 KTX recipients were available for mixed-effects modelling. At TP1, VP1 load was 21% (geometric mean, 95% CI 13%–35%). Time after diagnosis (start of the reduction of immunosuppression) was an independent predictor for VP1 decline: β = −0.044 ± 0.022, p = 0.047. Exponentiation of the coefficient indicated a 4% weekly decline in VP1-positive cells (expβ = 0.96, 95% CI 0.92–1.00). Raw observations are displayed in Figure 5A, while a population-level fitted curve with its 95% CI is shown in Figure 5B. To define a clinically meaningful threshold for VP1 positivity, we reclassified VP1 negativity as ≤1% to avoid trace analytical noise. We then examined VP1-positivity across predefined BKPyV-DNAemia bins. For the BKPyV-DNAemia in a range of 200–700 copies/mL, VP1 was almost always absent, with a median of 0% (IQR 0%–0.95% in 2 × 10²–4 × 10²; 0%–0.87% in 4 × 10²–5 × 10²) and positivity observed in only 3 of 12 cases (25%). By contrast, VP1 positivity increased substantially with higher viremia. At 7 × 10²–1 × 10³ copies/mL, 67% of samples were positive (median 2.65%, IQR 0.6–12.2), and at 1 × 10³–3 × 10³ copies/mL, 58% were positive (median 3.35%, IQR 0.9–12.9). Above 3 × 10³ copies/mL, VP1 positivity was frequent (3 × 10³–10 × 10⁴: 86%; 1 × 10⁴–1 × 10⁵: 88%), with median VP1 levels rising steeply (11.3% and 60.6%, respectively). These results indicate that VP1 positivity becomes unlikely once plasma BKPyV DNAemia falls into the ≤500–700 copies/mL range.