Urinary Biomarkers for Kidney Allograft Injury

Abstract

The current standard of serum creatinine and biopsy to monitor allograft health has many limitations. The most significant drawback of the current standard is the lack of sensitivity and specificity to allograft injuries, which are diagnosed only after significant damage to the allograft. Thus, it is of critical need to identify a biomarker that is sensitive and specific to the early detection of allograft injuries. Urine, as the direct renal ultrafiltrate that can be obtained noninvasively, directly reflects intrarenal processes in the allograft at greater accuracy than analysis of peripheral blood. We review transcriptomic, metabolomic, genomic, and proteomic discovery-based approaches to identifying urinary biomarkers for the noninvasive detection of allograft injuries, as well as the use of urine cell-free DNA in the QSant urine assay as a sensitive surrogate for the renal allograft biopsy for rejection diagnosis.

INTRODUCTION

For many patients with end-stage renal disease, dialysis and transplantation present opportunities to extend survival. Because kidney transplant recipients experience a superior quality of life and longer survival compared with those on dialysis,¹ transplantation presents the best treatment for end-stage renal disease. However, even with the development of immunosuppressant drugs, allograft rejection is a critical problem faced by ~10% of pediatric and adult kidney transplant recipients (KTRs).² The standard for monitoring of the allograft has remained mostly unchanged over the years, relying heavily on serum creatinine and biopsies to assess allograft function and health. Because serum creatinine increases late into injury and lacks specificity to allograft rejection, as it is confounded by tubular necrosis and BK virus nephropathy (BKVN), it does not present the necessary sensitivity and specificity as a biomarker for the early detection of kidney injury. Data have demonstrated that subclinical rejection, defined as rejection episodes not reflected by serum creatinine levels, can be identified in 35% of protocol biopsies within 2 y posttransplant.³ Furthermore, in small pediatric KTRs, significant rejection episodes may present as small changes in serum creatinine levels. Biopsies, as the gold standard for allograft rejection diagnosis, come with high cost, risk of bleeding, and potential observer bias. Addressing the shortcomings of serum creatinine and biopsies as means to monitor allograft health is a critical need that will improve the lives of 250 000 KTRs and 22 000 additional KTRs per year in the United States.² Thus, it is important to identify biomarkers that can be measured noninvasively and are more sensitive and specific than serum creatinine is to the early detection of rejection.

In recent years, biomarkers in blood have been the subject to many studies. However, the frequent need for in-person blood draws comes at the cost of inconvenience, risk, and in recent months, potential exposure to the coronavirus disease 2019 pandemic where social distancing is a current mandate but may continue to be a preference for the more immunocompromised transplant patient group even after resolution of the current pandemic. Noninvasive monitoring of kidney health via urine samples is greatly relevant to transplant patients who are likely at high risk for coronavirus disease 2019, as there is less physical contact with hospital staff compared with biopsy and collecting blood samples. Urine presents the ideal fluid to develop a robust biomarker for the early detection of allograft rejection, as it is the direct renal ultrafiltrate and can be collected noninvasively. The development of high-throughput techniques has paved the way for various omics approaches to identifying the ideal urinary biomarker. However, because of the challenges with degradation of various molecular components of urine, the investigation of biomarkers for kidney allograft health should be balanced with research regarding urine stability postcollection. Furthermore, studies have been performed to assess the risk of bias in kidney allograft rejection studies, highlighting the frequency of patient selection bias, the exclusion of the typical confounding of a real-life setting, risk of overfitting, and unrealistic performance for clinical applicability.⁴ In this review, we highlight various transcriptomic, genomic, metabolomic, and proteomic approaches to identifying urinary biomarkers for the early detection of kidney allograft rejection and injury and strategies to develop and validate biomarkers in urine for clinical applications in transplantation. Figure 1 demonstrates various omics approaches and associated techniques.

FIGURE 1.: Approaches for detecting kidney transplant clinical phenotypes. The figure above visually depicts the different technologies and methodologies for assessing important factors related to kidney transplantation—operational drug tolerance, allograft infection, acute rejection, and chronic injury. Approaches to understanding and diagnosing kidney transplant rejection and injury vary across different omic data sets. dd-cfDNA, donor-derived cell-free DNA; iKEA, integrated kidney exosome analysis; LC-Mass Spec, liquid chromatography with tandem mass spectrometry; miRNA, microRNA; mtcfDNA, mitochondrial cell-free DNA; RT-PCR, reverse transcription polymerase chain reaction; qPCR, quantitative polymerase chain reaction; SHERLOCK, Specific High-Sensitivity Enzymatic Reporter Unlocking assay; uCRM, urinary common rejection module.

Before a discussion about biomarkers for assessment of acute renal allograft rejection, it is important to review some basic tenants to put into context the evolution in our understanding about the mechanics and temporal nature of allograft rejection. Molecular profiling of the renal allograft during acute rejection (AR) has identified underlying molecular heterogeneity in AR over and above the heterogeneity observed by pathology between T cell–mediated rejection (TCMR) and antibody-mediated rejection (AMR).^5,6 Though the differentiation of rejection pathology and its causation is important for therapeutic choices and prognostication, it is also well recognized that both TCMR and AMR may coexist, that molecular injury may exist in histologically and clinically stable allografts,⁷ that AR is a “patchy” injury in the allograft, resulting in discordance between pathology and molecular changes, and that molecular changes in the allograft can precede both pathology and clinical features of rejection.^8,9 Furthermore, it is important to note that at this point, neither blood nor urine-based biomarkers provide clear distinction between TCMR and AMR.

In this review, rejection phenotypes that are not clearly defined as TCMR or AMR are considered as AR. Stable (STA) phenotypes refer to transplanted samples that are stable without confirmation of absence of rejection via biopsy. Healthy control (HC) phenotypes are transplanted control samples in which the absence of rejection or other phenotypes excluding rejection was biopsy confirmed.

URINARY BIOMARKERS: MECHANISTIC APPROACHES/BIOMARKER TYPES

Transcriptomics/RNA Biomarkers for AR, BKVN, and Tolerance in Urine

Transcriptomic hypotheses-generating studies of renal allograft rejection have identified potential biomarkers in mRNA, microRNA (miRNA), long noncoding RNA, and circular RNA for the early detection of allograft rejection. During TCMR, T cells infiltrate the allograft and measuring cellular levels of mRNA, particularly those targeted to cytotoxic T cells, has provided insights into T-cell immune responses in rejection, inclusive of CD3ε, perforin, granzyme B, proteinase inhibitor 9, CD103, IP-10, CXCR3, and granulysin.^10,11 A 3-gene TCMR diagnostic mRNA signature (CD3ε, IP-10, and the housekeeping gene 18S rRNA) was identified from 4300 urine samples, and detected biopsy confirmed TCMR from non-TCMR samples with 79% sensitivity and 78% specificity.¹¹ Furthermore, through profiling urinary cell mRNA of 26 AMR, 26 TCMR, and 32 acute tubular injury samples, the same group identified a 5-gene signature (CD3ε, CD105, CD14, CD46, and 18S rRNA) that differentiated TCMR from AMR.¹² Similarly, a set of 11 mRNAs in urine expanded on a previous public and private genomic meta-analysis study of over 700 kidney transplants that demonstrated that 11 common rejection module (CRM) genes in solid organ transplant rejection tissue are overexpressed in AR patients irrespective of type of tissue organ (heart, lung, liver, or kidney transplant).^13,14 These 11 CRM genes were also evaluated in urine (uCRM) by Sigdel et al,¹⁵ as BASP1, CD6, CXCL10, CXCL9, INPP5D, ISG20, LCK, NKG7, PSMB9, RUNX3, and TAP1, across 150 unique urine samples. All but RUNX3 transcripts were significantly elevated in rejection samples compared with those without rejection, and expressions of CD6, CXCL10, CXCL9, NKG7, and PSMB9 were also significantly increased in borderline AR (bAR) samples, defined as the infiltration of mononuclear cells (<25% of the parenchyma) or foci of mild tubulitis (1–4 mononuclear cells/tubular cross-section), compared with HC samples. After accounting for the differential expression of the CRM gene set across rejection and stable groups through nonlinear supervised methods, a uCRM score was developed across all 11 genes. The uCRM score differentiated AR from HC with 95.35% sensitivity and 97.78% specificity at a threshold of 3.60. At the same threshold, the uCRM score differentiated AR from the combination of bAR and HC with 87.10% sensitivity and 97.78% specificity and differentiated AR from the combination of BKVN, bAR, and HC with 76.92% sensitivity and 97.78% specificity. Given the functional roles for many of these genes in the immune response, the uCRM score demonstrated a significant correlation with the tubulitis and interstitial inflammation biopsy scores in AR, based on contemporaneous biopsy histology and urine uCRM sampling in this data set. Such correlation highlights potential utility of uCRM as a transplant monitoring tool, specifically to screen patients before protocol or indication biopsies. However, although uCRM and other mRNA-based tests indeed provide utility as a monitoring tool, mRNA degradation because of poor sample processing, or lengthy periods between collection and processing, often results in sample exclusion. As a result, proper quality control checks and other quality management systems should be in place to prevent such nucleic acid deterioration and rejection of samples.

Kaminski et al¹⁶ have developed a CRISPR-Cas13-based diagnostic test for the detection of BKV and AR. The authors used a modified version of the CRISPR/Cas13–specific high-sensitivity enzymatic reporter unlocking (SHERLOCK) assay to detect BKV. Briefly, regions of 95% homology across all strains of each virus were amplified via recombinase polymerase amplification, and primer pairs and CRISPR RNAs were then tested for their abilities to detect American Type Culture Collection diagnostic BKV standard. The modified SHERLOCK assay identified all BKV urine samples (15 BKV positive, 16 BKV negative) with 100% sensitivity and specificity with and without the heating unextracted diagnostic samples to obliterate nucleases (HUDSON) protocol. Furthermore, the authors explored whether SHERLOCK could be used to detect CXCL9 mRNA levels to indicate AR. Reverse transcription-recombinase polymerase amplification was used to amplify the CXCL9 mRNA, and Cas13 alone was sufficient to detect CXCL9. Across 14 AR and 17 HC samples, the CXCL9 assay detected AR with a 93% sensitivity and 76% specificity. The authors then combined the HUDSON protocol with SHERLOCK and lateral-flow dipsticks to develop a rapid point-of-care test. This test, paired with a smartphone application, could output a binary positive/negative result based on the lateral-flow strip picture taken via a smartphone camera. Clinically, in a kidney transplant recipient admitted for graft dysfunction, the lateral flow–based CRISPR diagnostic detected BKV [confirmed by quantitative polymerase chain reaction (qPCR) and biopsy]. The CRISPR-based assay still requires more work regarding optimizing mRNA isolation and testing in a larger prospective study, and the next step will be investigating use of this assay in detecting AMR.

In addition to mRNAs, miRNAs have been studied as urinary biomarkers for allograft rejection. Lorenzen et al¹⁷ used qPCR to identify 3 deregulated miRNAs, miR-10a, miR-10b, and miR-210, in AR samples. In a validation cohort of 62 AR, 13 STA with urinary tract infections, and 19 HC samples, AR samples were found to have significantly decreased miR-10b and miR-210 levels and significantly elevated miR-10a levels compared with HCs. miR-210 diagnosed AR from HC with 74% sensitivity and 52% specificity and was significantly associated with renal function decline (based on the estimated glomerular filtration rate [GFR]) 1 y posttransplantation. Furthermore, Millán et al¹⁸ used qPCR on 80 samples (8 ARs, 72 HCs) to identify miR-142-3p, miR-210-3p, and miR-155-5p as significantly deregulated miRNAs during AR. miR-155-5p was identified as the optimal miRNA for the diagnosis of AR, as it diagnosed AR with 85% sensitivity and 86% specificity [88% positive predictive value (PPV), 100% negative predictive value (NPV)]. CXCL10 was consistently elevated in AR samples compared with HC, and CXCL10 mRNA could identify AR with 84% sensitivity and 80% specificity (90% PPV, 85% NPV). Both miR-155-5p and CXCL10 mRNA were significantly correlated with the estimated GFR and returned to normal levels after antirejection therapy. Other studies have also explored the use of long noncoding RNAs and circular RNAs as biomarkers for allograft rejection.^19-21

In addition to evaluating biomarkers to detect rejection and allograft injury, investigators have also evaluated a rare group of patients who can develop prope or operational tolerance to their renal allograft after deliberate immunosuppression (IS) withdrawal because of nonadherence or prior problems with over-IS (infection or malignancy). It is possible that urine biomarkers indicative of this allograft tolerance state may help to identify patients who can have a low risk of rejection and tolerate minimization of IS therapy. Though the vast majority of research regarding biomarkers for tolerance has been established on blood samples,^22,23 urine samples from a study of induced tolerance via a combined bone marrow and kidney transplantation strategy in 23 samples (6 STA without IS therapy [tolerant], 8 STA on IS therapy [STA-IS], 5 nontransplanted normal, 4 AR) demonstrated increased levels of Foxp3 mRNA in tolerant samples compared with STA-IS or nontransplanted normal kidney samples, and decreased levels of Foxp3 mRNA in tolerant samples compared with AR samples.²⁴ Furthermore, across urine samples from the study by Newell et al²³ that evaluated blood-based markers for tolerance with 25 tolerant, 33 STA-IS, and 42 nontransplanted normal samples, urinary cell mRNA levels for CD20, Foxp3, CD3ε, and perforin mRNA were increased in tolerant samples compared with nontransplanted normal samples. CD20 was additionally found in increased levels in tolerant samples compared with STA-IS samples. Foxp3 mRNA was not statistically elevated in tolerant samples compared with STA-IS, contradicting the results from a previous study of the same group. The authors¹² further investigated the urinary levels of T cell–associated transcripts in urine from 5 tolerant and 5 nontolerant samples (3 subclinical rejection and 2 kidney disease recurrence) where tolerance was induced via donor hematopoietic stem cells, alemtuzumab induction, initial drug therapy with tacrolimus/mycophenolate IS and conversion to sirolimus, and drug withdrawal 24 mo after transplantation.²⁵ Reverse transcription quantitative polymerase chain reaction revealed that mean levels of 19 measured mRNAs were lower in tolerant compared with nontolerant samples. Notably, mean levels of Foxp3, perforin, and CD20 were decreased in tolerant samples. The main limitations of these studies are the small sample size, which is difficult to overcome because of the small number of tolerant patients, and the likely varying mechanisms of tolerance acquisition in the different protocols. Furthermore, inconsistent results such as Foxp3 and CD20 mRNA levels in tolerant samples compared with STA-IS samples suggest that some of these changes in this comparative expression analysis between operational/induced tolerance and nontolerant patients on IS reflect the inherent gene expression changes driven from IS exposure rather than the tolerant state itself. Hence, application of these markers for any IS titration has not been successfully attempted to date.

Metabolomics/Metabolite Biomarkers for AR and Graft Injuries in Urine

Metabolite analysis using liquid chromatography–mass spectrometry (LC-MS) on 1516 urine supernatant samples from 241 KTRs identified 3-sialyllactose, xanthosine, quinolinate, and X-16397 as metabolites that could diagnose TCMR from non-TCMR samples in a study by Nissaisorakarn et al.¹² When the metabolites were combined with the 3-gene signature developed in the same multicenter study (CTOT-04) by the same group, the resulting signature diagnosed TCMR with 90% sensitivity and 84% specificity in a selected cohort of 39 TCMR and 159 non-TCMR samples. The improved performance of the previously developed CTOT-04 3-gene signature upon the addition of metabolites in diagnosing TCMR highlights the benefits of using a multiomics signature for urine biomarker panels. Sigdel et al²⁶ used gas chromatography–mass spectrometry across a large cohort of patients with biopsy-confirmed urine samples: 106 AR, 111 HC, 71 interstitial fibrosis and tubular atrophy, and 22 BKVN. The authors identified 266 metabolites, further selected down to a 9-metabolite model (glycine, N-methylalanine, adipic acid, glutaric acid, inulobiose, threitol, isothreitol, sorbitol, and isothreonic acid) for identification of AR. The 9-metabolite model identified AR or interstitial fibrosis and tubular atrophy from HC with 95.3% sensitivity and 75.9% specificity. Across 22 BKVN and 288 non-BKVN samples, 5 metabolites (arabinose, 2-hydroxy-2-methylbutanoic acid, hypoxanthine, benzyl alcohol, and N-acetyl-d-mannosamine) were identified to detect BKVN with 72.7% sensitivity and 96.2% specificity. In 22 BKVN and 111 HC samples, 4 metabolites (arabinose, 2-hydroxy-2-methylbutanoic acid, octadecanol, and phosphate) were identified to detect BKVN with 88.9% sensitivity and 94.8% specificity. Moreover, across 106 AR and 111 HC samples, 11 metabolites (glycine, adipic acid, glutaric acid, N-methylalanine, inulobiose, threose, sulfuric acid, taurine, asparagine, 5-aminovaleric acid, and myoinositol) were identified to detect AR with 92.9% sensitivity and 96.3% specificity [area under the curve (AUC) 98.5]. This study highlights that panels of urine metabolites may have the ability to detect specific transplant injuries with high accuracy. Blydt-Hansen et al²⁷ explored the use of 134 urine metabolites, identified via MS, in 30 TCMR, 183 non-TCMR (excluding BKVN), and 54 borderline tubulitis samples across pediatric KTRs. Using partial least squares discriminant analysis, the authors identified 10 metabolites inclusive of proline, kynurenine, sarcosine, methionine sulfoxide, threonine, glutamine, phenylalanine, and alanine, which identified TCMR with 83% sensitivity and specificity (97% NPV and 45% PPV). When 54 borderline tubulitis samples were compared with the no TCMR group, the authors found 5 metabolites that overlapped with those selected in the TCMR model. When the TCMR model was applied to borderline tubulitis samples, the discriminant score was midway between TCMR and no TCMR; the same relationship was observed when the borderline tubulitis model was applied to TCMR samples, suggesting that allograft injury from TCMR exists on a continuum of severity. Thus, the authors combined TCMR and borderline tubulitis to compare with no TCMR samples. In the validation cohort, TCMR/borderline tubulitis was distinguished from no TCMR samples with 65% sensitivity and 74% specificity. In all, these findings from various metabolomic studies demonstrate the potential use for metabolites in TCMR detection. However, the main limitations of a pure metabolomic signature reside in the high cost and limited availability of MS. As demonstrated by Sigdel et al and Blydt-Hansen et al, a large metabolite signature is necessary to achieve high diagnostic accuracy, and such a large signature may be a financial and logistic burden regarding clinical application. Combining metabolite biomarkers with other omics markers can pave the way for an AR diagnostic signature that can be efficiently translated to clinical application. This application is discussed further in this Review (see Combined Omics).

Genomics/DNA Biomarkers of AR in Urine

Zhang et al²⁸ used PCR to explore the use of donor-derived cell-free DNA (dd-cfDNA) as a biomarker for AR. SRY and DYZ-1 genes were selected to identify dd-cfDNA in female recipients of male donors. During AR episodes (n = 4), SRY and DYZ-1 were detected, and after successful antirejection therapy (n = 3), SRY and DYZ-1 were not detected. It is important to note that 1 sample was SRY⁺ among 14 STA and 2 samples were DYZ-1⁺ among 12 STA samples. In addition, HLA-DRB1 was selected to identify dd-cfDNA in recipients of HLA-mismatched allografts. HLA-DRB1 was detected in 16 of 18 AR samples, and of the 16, 13 were HLA-DRB1^– after successful antirejection therapy. Across 23 STA samples, there were 2 HLA-DRB1⁺ samples; these observations suggest that dd-cfDNA may reflect different graft injuries and not just graft AR. Zhong et al²⁹ used nested PCR and reverse transcription polymerase chain reaction on 25 female recipients of male kidneys. ChrY dd-cfDNA was elevated in 3 AR patients, and after successful antirejection therapy, dd-cfDNA declined to nearly undetectable levels. In addition to monitoring dd-cfDNA levels in the 3 AR patients across 30 d posttransplantation, the authors observed elevated levels of dd-cfDNA in a patient who experienced rejection many months after transplantation. Sigdel et al³⁰ similarly explored the use of ChrY dd-cfDNA as a biomarker for allograft injury and rejection using digital PCR. Digital PCR was used to measure ChrY dd-cfDNA levels in female recipients of male donors. Across 63 biopsy-matched samples (41 HC, 8 TCMR, 10 chronic allograft injury [CAI], and 4 BKVN), dd-cfDNAs in TCMR and BKVN samples were significantly elevated when compared with HC and CAI, but the difference between TCMR and BKVN samples was not found to be significant. Thus, to expand the use of dd-cfDNA for AR diagnosis, further biomarkers may be needed to address the confounding factors of infection and different types of allograft injury. Additionally, as dd-cfDNA displayed significant correlation with protein/creatinine ratio, GFR, and acute allograft injury, the increased levels of urinary dd-cfDNA may reflect the increased burden of tissue injury and apoptosis. It is important to note that the approach of measuring ChrY dd-cfDNA in these 3 studies is limited to female recipients of male allografts. Thus, in cases whereas ChrY dd-cfDNA cannot be measured, the donor DNA would need to be sequenced.

Using a different approach, Kim et al³¹ explored the use of urinary mitochondrial cell-free DNA (mtcfDNA) as a biomarker for AR and delayed graft function (DGF) in KTRs, where mtcfDNA was measured using qPCR. Across 85 urine samples, the authors found significantly elevated mtcfDNA levels in AR (n = 12) compared with HC. However, because of the small sample size and a short follow-up period, and the complexity of urine sample optimization for sequencing and the overall costs of sequencing, the performance of mtcfDNA in diagnosing AR was not further investigated. The authors further investigated the association between urinary mtcfDNA and graft function, where DGF was characterized as needing dialysis within 7 d posttransplant and slow graft function was defined as a <20% decrease in serum creatinine levels within 24 h posttransplant. Urinary mtcfDNA levels were significantly elevated in KTRs with slow graft function and DGF compared with those with immediate graft function, defined as a >20% decrease in serum creatinine levels 24 h posttransplant. Regression modeling demonstrated a statistically significant positive correlation between mtcfDNA and posttransplant renal recovery time and a significantly better graft outcome for KTRs with low mtcfDNA levels. This study demonstrated the association of mtcfDNA with graft injury and function early after transplantation, but additional studies are needed to validate these findings. In a study regarding total plasma cfDNA as a biomarker for AR, Moreira et al³² observed an increase in urinary cfDNA during rejection episodes and a decrease following antirejection treatment. Increased levels of cfDNA were also observed in cases of urinary sepsis, once again highlighting the nonspecificity of urinary cfDNA alone in diagnosing AR. Nevertheless, DNA biomarkers may present a valuable complement to other biomarkers for the early detection of rejection (see Combined Omics).

Proteomics/Protein and Peptide Urine Biomarkers for AR and Graft Injury

Park et al³³ developed urine-based integrated kidney exosome analysis (iKEA) platform to detect TCMR. The assay immunomagnetically captures target exosome vesicles (EVs), which are then labeled with an oxidizing enzyme through a secondary antibody. The electrical current generated by the enzymatic reaction is measured by the portable iKEA detector that can transfer data via Bluetooth or universal serial bus. The iKEA platform was applied to 15 TCMR and 15 non-TCMR urine samples, and CD3, CD45, CD2, HLA-ABC, and CD52 were selected as T-cell markers as they were highly expressed in urinary EVs. CD3⁺ EVs were significantly elevated in the TCMR samples, whereas CD3⁺ EVs were measured at low levels in BKVN and AMR samples. In the discovery phase, the iKEA assay detected TCMR with 92.8% sensitivity, 87.5% specificity, and 90% accuracy, and during validation, detected TCMR with 63.6% sensitivity, 100% specificity, and 71.4% accuracy. The main limitation of this study was the small sample size (n = 30 for discovery, n = 14 for validation). Sigdel et al³⁴ applied LC with tandem MS (LC-MS/MS)–based shotgun proteomics to identify urinary biomarkers for biopsy-confirmed AR. Across 40 samples (10 AR, 10 HC, 10 nontransplant proteinuria, and 10 nontransplant control), 9 proteins were identified only in AR samples, and Tamm-Horsfall protein (UMOD), Pigment Epithelium-Derived factor (PEDF) or SERPINF1, and CD44 were further selected for verification by ELISA. The 3 proteins identified AR independent of confounding variables such as proteinuria, IS, age, and sex, and CD44, PEDF, and UMOD displayed 0.973 AUC, 0.932 AUC, and 0.846 AUC, respectively. In an additional study,³⁵ the same group analyzed 245 urine samples (112 AR, 117 HC, 116 CAI, and 51 BKVN) via LC-MS using isobaric tags for relative and absolute quantitation reagents, discovering 517 proteins that were significantly altered in AR samples, 186 proteins specific to CAI, and 108 proteins specific to BKVN. The proteins were narrowed down to 296 peptides mapping to 100 proteins for selected reaction monitoring validation, which identified 11 peptides specific to AR (0.939 AUC), 12 specific to CAI (0.995 AUC), and 12 specific to BKVN (0.832 AUC) when compared with HC in an independent cohort (n = 151).

Lim et al³⁶ investigated the use of exosomal proteins to detect TCMR across 22 HC and 25 TCMR-confirmed urine samples. Upon analysis by LC-MS/MS of exosomes, the authors identified tetraspanin-1 and hemopexin as proteins that were significantly elevated in TCMR samples, with 64% sensitivity and 72.7% specificity. This study was the first to identify proteomic biomarkers in an Asian population albeit based on a small sample size. Kanzelmeyer et al³⁷ studied the use of urinary proteomics to diagnose chronic AMR in pediatric patients with 24 chronic AMR (cAMR) and 36 non-AMR samples analyzed via capillary electrophoresis-mass spectrometry, identifying 79 peptides associated with cAMR. After adding CKD273 as an additional biomarker, the classifier diagnosed cAMR with 88% sensitivity and 92% specificity. Given the fact that the cAMR samples all had advanced chronicity, the study results will need to be validated that these biomarkers are not being skewed by the chronic background injury in the allograft and that the biomarkers will also display similar performance in an adult cohort. Mertens et al³⁸ have also explored the use of urinary proteomics to diagnose AMR. Shotgun LC-MS/MS on 60 AMR and 189 non-AMR samples identified 10 proteins (alpha-1-B glycoprotein, afamin, apolipoprotein A1, apolipoprotein A4, Ig heavy constant α1, Ig heavy constant γ4, leucine-rich α2-glycoprotein 1, alpha-1 antitrypsin, antithrombin, and transferrin) that could diagnose AMR with 95% sensitivity and 96% specificity. When the 10-protein model was tested on an independent validation cohort of 43 AMR and 348 non-AMR samples, it detected AMR with 95% sensitivity and 76% specificity (0.88 AUC). The 10-protein model displayed significant association with lesions of AMR; however, it also displayed significant association with BKVN and glomerulonephritis and classified around 50% of TCMR cases as AMR cases. The authors attribute such correlation to the fact that the 10 proteins were selected on the basis of statistics rather than biology, but the limited specificity of the 10-protein panel brings its clinical application into question, as further testing such as a biopsy would be required in case of positive test results. Hirt-Minkowski et al³⁹ used sandwich ELISA across 213 KTRs to detect tubulointerstitial allograft inflammation. The authors used Banff scores to classify allograft histology results into no relevant inflammation (acute score zero and the interstitial infiltrates) and inflammation (tubulitis t1, tubulitis t2–3, and isolated vascular compartment inflammation) groups. Receiver operating characteristic analysis on 362 surveillance biopsies at 3 and 6 mo posttransplant (243 no inflammation, 119 inflammation) demonstrated that CXCL10 detected subclinical inflammation with 61% sensitivity and 72% specificity (0.69 AUC). Furthermore, ROC analysis on 80 indication biopsies (50 no inflammation, 30 inflammation) demonstrated that CXCL10 detected clinical inflammation with 63% sensitivity and 80% specificity (0.74 AUC). The authors found that using CXCL10 to monitor allograft health would have reduced surveillance biopsies by 61% and indication biopsies by 64%. This was the first study to investigate the use of a noninvasive biomarker to monitor allograft health in a real-life setting. More recently, Raza et al⁴⁰ used ELISA to quantify urinary CXCL10 protein levels in 96 AR and 89 non-AR samples. Urinary CXCL10 protein levels detected AR with 72% sensitivity and 71% specificity (0.74 AUC). More specifically, CXCL10 detected TCMR (n = 63) from non-AR samples with 79% sensitivity and 71% specificity. Furthermore, the authors demonstrated a significant difference in rejection-free graft survival time between CXCL10 low (<100 pg/mL) and high (>200 pg/mL) protein levels. The findings from this study are in line with various studies that have demonstrated CXCL10 as a promising biomarker for allograft rejection,^41-43 but additional biomarker combinations beyond CXCL10 alone will likely be needed to improve rejection detection accuracy. Jackson et al⁴³ studied urinary CXCL9 and CXCL10 levels across pediatric and adult samples inclusive of 25 AR, 24 BKV, 50 HC, and 31 nontransplant control samples. Both CXCL9 and CXCL10 were significantly elevated in AR and BKV samples, whereas their levels were not different between the AR and BKV groups. CXCL9 detected either AR or BKV with 86% sensitivity and 80% specificity and CXCL10 detected either AR or BKV with 80% sensitivity and 76% specificity. The authors additionally demonstrated that CXCL9 and CXCL10 can detect subclinical rejection and BKV. Furthermore, the authors showed that urinary chemokine measurements were consistent up to 6 freeze-thaw cycles, highlighting the stability of CXCL9 and CXCL10 as biomarkers for monitoring allograft health. Of the peptides and proteins discussed above, CXCL9 and CXCL10 are promising biomarkers for the diagnosis of AR, as they are well studied relative to other proteomic markers and can be measured via ELISA rather than MS, which is expensive and restricted in availability. Nevertheless, the elevations of these urinary chemokines may detect different inflammatory injuries in the allograft, over and above AR alone. Consequently, when using such chemokine markers, Tinel et al⁴⁴ note that risk-stratification models may be more useable in clinical situations and could allow for better determination of disease cause.

Combined Omics/DNA, Protein, and Metabolite Urine Panels for AR

As reviewed above, all individual measurements of different biomarker modalities in urine provide informative inferences about diagnosis of AR when paired with a contemporaneous biopsy that confirms histological rejection. Nevertheless, the accuracies of these biomarker panels (as shown in Table 1) are either based on small sample sizes, the absence of matched biopsy data for all profiled urine samples, insufficient samples with both TCMR and AMR in the same data set, or performance metrics only in specific populations such as adult or pediatric. In addition, RNA extraction from the urine pellet is biased with very fragmented and low-quality RNA and is not accompanied by a fail-safe guarantee of RNA extraction and quality success. Urine also presents challenges with being confounded by urinary tract injuries, the rapid degradation of different molecules because of the variations in pH, the intrinsic activation of proteases and nucleases (more so in transplant injury), contaminants and infectious pathogens, and the processing steps required to stabilize and process urine optimally. Despite these challenges, recent urine assays for transplant rejection may provide the safest, most convenient, rapid, and cost-effective pathway for optimizing the posttransplant IS and management of all renal allograft recipients. This has important ramifications for improving adherence by repeated monitoring and is highly attractive for noninvasive immune risk assessment in young children, where the invasive biopsy carries greater risks because of associated risks of accompanying conscious sedation and general anesthesia.

TABLE 1. - Summary of relevant urine biomarkers for renal allograft injury

Reference	Biomarkers	Test design	Sensitivity	Specificity	PPV	NPV	AUC
Transcriptomics
Suthanthiran et al11	CD3ε mRNA + IP-10 mRNA + 18S rRNA	TCMR vs non-TCMR	79%	78%	–	–	0.850
Nissaisorakarn et al12	CD3ε + CD105 + CD14 + CD46 + 18S rRNA	TCMR vs AMR	–	–	–	–	0.810
Sigdel et al15	BASP1, CD6, CXCL10, CXCL9, INPP5D, ISG20, LCK, NKG7, PSMB9, RUNX3, TAP1 (uCRM Score)	AR vs HC	95.35%	97.78%	–	–	0.9886
		AR vs bAR + HC	87.10%		–	–	0.9677
		AR vs BKVN + bAR + HC	76.92%		–	–	0.9111
Kaminski et al16	CXCL9 mRNA Unnecessary	AR vs HC	93%	76%	–	–	0.91
Lorenzen et al17	miR-210	AR vs HC	74%	52%			0.70
Millán et al18	miR-155	AR vs non-AR	85%	86%	88%	100%	0.875
	CXCL10 mRNA		84%	80%	90%	85%	0.865
Metabolomics
Nissaisorakarn et al12	3-sialyllactose, xanthosine + quinolinate + X-16397 + CD3εmRNA + IP-10 mRNA + 18S rRNA	TCMR vs non-TCMR	90%	84%	–	–	0.930
Sigdel et al26	Glycine, adipic acid, glutaric acid, N-methylalanine, inulobiose, threose, sulfuric acid, taurine, asparagine, 5-aminovaleric acid, myoinositol	AR vs HC	92.9%	96.3%	96.3%	92.9%	0.985
	arabinose, 2-hydroxy-2-methylbutanoic acid, octadecanol, and phosphate	BKVN vs non-BKVN	88.9%	94.%	72.7%	98.2%	0.940
Blydt-Hansen et al27	Proline, PC:aa:C34:4, kynurenine, sarcosine, methionine sulfoxide, PC:ae:C38:6, threonine, glutamine, phenylalanine, alanine	TCMR vs non-TCMR	83%	83%	97%	45%	0.880
		TCMR + borderline tubulitis vs non-TCMR	95% (training) 74% (validation)	75% (training) 65% (validation)	–	–	0.900 (training)
Proteomics
Park et al33	CD3+ extracellular vesicles (iKEA)	TCMR vs non-TCMR	92.8% (discovery) 63.6% (validation)	87.5% (discovery) 100% (validation)	–	–	0.911 (discovery) 0.837 (validation)
Sigdel et al34	Tamm-Horsfall protein (UMOD)	AR vs non-AR	–	–	–	–	0.973
	Pigment Epithelium-Derived factor (PEDF) or SERPINF1		–	–	–	–	0.932
	CD44		–	–	–	–	0.846
Sigdel et al35	11-peptide panel	AR vs HC	–	–	–	–	0.939
	12-peptide panel	BKVN vs HC	–	–	–	–	0.832
	12-peptide panel	CAI vs HC	–	–	–	–	0.995
Lim et al36	Tetraspanin-1 and hemopexin	TCMR vs HC	64%	72.%	–	–	0.744
Kanzelmeyer et al37	79-peptide panel	cAMR vs non-cAMR	100%	75%	–	–	0.92
	79-peptide panel + CKD273		88%	92%	–	–	0.92
Mertens et al38	Alpha-1-B glycoprotein, afamin, apolipoprotein A1, apolipoprotein A4, Ig heavy constant α1, Ig heavy constant γ4, leucine-rich α2-glycoprotein 1, alpha-1 antitrypsin, antithrombin, and transferrin	AMR vs non-AMR	95% (training) 95% (validation)	96% (training) 76% (validation)	–	–	0.98 (training) 0.88 (validation)
Hirt-Minkowski et al39	CXCL10 protein	Inflammation vs no inflammation	61% (surveillance biopsy) 63%(indication biopsy)	72% (surveillance biopsy) 80%(indication biopsy)	–	–	0.69 (surveillance biopsy) 0.74(indication biopsy)
Raza et al40	CXCL10 protein	AR vs non-AR	72%	71%	–	–	0.74
		TCMR vs non-AR	79%	71%	–	–	0.79
Jackson et al43	CXCL9 protein	AR or BKV vs CNI toxicity + IFTA + HC + nontransplant control	86%	80%	–	–	–
	CXCL10 protein	AR or BKV vs CNI toxicity + IFTA + HC + nontransplant control	80%	76%	–	–	–
Combined omics
Yang et al45	Multiple biomarker types: cfDNA, m-cfDNA, CXCL10, creatinine, clusterin, total protein (Q Score/QSant)	AR vs HC	94.9% (training) 95.8% (validation)	100% (training) 99.3% (validation)	–	–	0.99 (training) 0.998 (validation)

The table displays the ROC analysis results of various methods to identify allograft injury in urine samples. The bold text in parentheses indicate assay names.

AMR, antibody-mediated rejection; cAMR, chronic AMR; AR, acute rejection; AUC, area under the curve; bAR, borderline acute rejection; BKV, BK virus; BKVN, BKV nephropathy; cfDNA, cell-free DNA; CNI, calcineurin inhibitor; HC, healthy control; HUDSON, heating unextracted diagnostic samples to obliterate nucleases; IFTA, interstitial fibrosis and tubular atrophy; iKEA, integrated kidney exosome analysis; NPV, negative predictive value; m-cfDNA, microbial cell-free DNA; PPV, positive predictive value; ROC, receiver operating characteristic; TCMR, T cell–mediated rejection.

To address these multiple confounders in the performance of a urine biomarker for transplant rejection, Yang et al⁴⁵ have developed a custom urine assay consisting of a chemiluminescent immunoprobe to measure cell-free DNA and its epigenetic changes in urine supernatant, without the need for RNA extraction, amplification, or sequencing. Yang et al have developed a composite artificial intelligence–based algorithm using specific demographic variables, as well as additional protein (total protein, CXCL10, clusterin) and metabolite (creatinine) biomarkers on 601 urine samples, with the ability to normalize results in a manner that the assay can be done on a randomly collected spot urine sample. The performance of the assay, called QSant (finding “clues about transplant health”), has high accuracy to detect AR in kidney transplant patients.^45,46 The biomarkers in the QSant assay compute to generate a scaled Q Score from 0 to 100, developed on a training set of 39 AR and 72 HC samples, 2 independent validation sets, and a prediction set that can detect AR up to 200 d before rejection diagnosis by biopsy. Q Scores <32 corresponded to no active rejection and Q Scores ≥32 indicated an increased risk of active rejection. At the 32 Q Score threshold, the QSant assay diagnosed AR with 94.9% sensitivity and 100% specificity (0.99 AUC). Upon further validation across 385 samples, the QSant assay achieved 95.8% sensitivity and 99.3% specificity (0.998 AUC). The Q Score did not differ between TCMR and AMR cases and diagnosed TCMR and AMR with equally high accuracy in both children and adults. Additionally, the Q Scores did not significantly differ between subclinical AR (n = 32) and clinical AR (n = 39) samples, demonstrating accuracy of QSant in diagnosing subclinical AR when graft function is at baseline. QSant demonstrates the excellent sensitivity, specificity, and AUC in detecting AR and may provide a monitoring surrogate approach for serial immune risk surveillance after kidney transplantation.

FUTURE DIRECTIONS

Highly accurate monitoring methods are needed to supersede the current practice of tracking serum creatinine drifts only as a means of graft dysfunction-driven biopsies for rejection diagnosis, as this is too late to change outcomes and extend graft survival trajectories. Urine biomarkers provide great promise for serial, noninvasive monitoring for the surveillance of acute renal allograft rejection and for evaluation of patients before protocol biopsies to reduce procedure waste, to predict early rejection with the opportunity to reverse rejection by earlier IS intensification, and to monitor for complete resolution of inflammation and rejection. Moving forward, additional prospective clinical trials using the assays discussed in this review will further refine the performance of these assays, particularly the PPV and NPV values, which may be too optimistic in cross-sectional studies that may have sample selection bias. Furthermore, a combined omics approach to identifying biomarkers for allograft injury detection is promising, as the analysis of molecules from different molecular levels may indicate specific signaling pathways and networks. Another area of focus is to identify biomarkers to distinguish TCMR from AMR, as at this point, neither blood nor urine-based markers provide a clear distinction between TCMR and AMR. The group involved in the QSant assay is currently developing a new registry study that will allow for further validation in a multicenter prospective clinical trial. The large registry will not only provide additional data on performance and accuracy of QSant but also support future registry-based clinical trials.

CONCLUSION

A major unanswered question in the field of transplant diagnostics is whether the renal allograft biopsy is indeed a true gold standard for diagnosing allograft injury and if it is a sufficient standard for conclusive diagnosis of renal allograft health and immune quiescence. Recent molecular studies on the allograft and advanced urine bioassays that provide high correlation with allograft inflammation and tubulitis suggest that the allograft biopsy may actually be the “copper-standard” for a stable allograft, where clinical and histological injuries may lag behind molecular fingerprints of otherwise “hidden” graft rejection. The evolution of multiomic assays in urine may forge new clinical protocols for personalizing IS by increased frequency of noninvasive, serial urine monitoring for allograft health and injury.