Validation of a novel non-invasive method based on evaluating urinary exfoliated proximal tubule kidney cell multispectral autofluorescence to differentiate between causes of kidney transplant complications

Introduction: Exfoliation of proximal tubule cells (PTCs) into urine is a physiological process which may generate significant information regarding an individual's kidney status, with PTCs forming majority of the kidney mass. To date, there remain no validated method apart from transplant biopsy to accurately identify between histopathological causes of transplant graft dysfunction. Cell multispectral autofluorescence is sensitive to metabolic changes and oxidative stress, key processes driving kidney pathological manifestations. Considering an unmet critical need for non-invasive diagnostic approaches to reliably identify kidney transplant complications, our group proposed and subsequently patented a novel non-invasive methodology (AU2024904153) based on the assessment of multispectral autofluorescence properties in extracted urinary exfoliated PTCs from kidney transplant recipients. We conducted a pilot study applying this methodology, where binary classification of urinary exfoliated PTCs based on multispectral autofluorescence comparing between study participants with acute tubular necrosis (ATN) vs graft rejection (GR), with ATN vs interstitial fibrosis & tubular atrophy (IFTA), and with GR vs IFTA displayed excellent ROC AUC values of 0.95, 0.92 and 0.91 respectively.¹ We now aim to reproduce these results in a validation cohort.
Methods: 57 patients diagnosed with ATN (19 patients), GR (19 patients) or IFTA (19 patients) by transplant graft biopsy were included. Exfoliated PTCs were extracted using a validated immunomagnetic separation methodology. Cell imaging was performed on a multispectral autofluorescence microscopy and camera system. The optimal sets of autofluorescence feature combinations which significantly discriminate between the groups when comparing ATN vs GR, ATN vs IFTA, and GR vs IFTA were identified. Binary classification was performed by the AutoGluon machine-learning software.
Results: There were no statistically significant differences in demographic and clinical characteristics between the three groups. Multispectral autofluorescence imaging was completed for 238 PTCs in total (81 PTCs in the ATN group; 71 PTCs in the GR group; 86 PTCs in the IFTA group). Applying our methodology, binary classification of urinary exfoliated PTCs based on multispectral autofluorescence when comparing between study participants with ATN vs GR, with ATN vs IFTA, and with GR vs IFTA displayed ROC AUC values of 0.91, 0.89 and 0.93 respectively (Figure).

Conclusion: These results in a validation cohort supported findings from our initial pilot study¹, that machine-learning based assessment of urinary exfoliated PTCs multispectral autofluorescence could determine between ATN, GR and IFTA in kidney transplant recipients to an excellent degree of accuracy. Our novel methodology demonstrates immense potential as a non-invasive tool to guide post-transplant management going forward.

¹Wu HHL et al., Kidney360. 2025;6(11):1853-1862.