4.1 Borderline Changes and SCR

Borderline changes represent the lowest detectable indicator of TCMR histology [35]. However, the definition and significance of borderline changes as they relate to SCR are not straightforward [41]. The original 1997 Banff assessment defined “borderline changes suspicious for rejection” as tubulitis (Banff “t”) with at least modest interstitial inflammation (Banff “i”) [42]. The definition of borderline changes was broadened in 2005–2007 to reduce or eliminate the need for the finding of interstitial inflammation. This defined borderline changes as foci of tubulitis only [43]. The 2019 Banff report reestablished the classification of “borderline changes” as: interstitial inflammation involving 10%–25% of nonsclerotic cortex (Banff i1) with at least mild tubulitis (t  > 0) [44]. It is therefore important that careful attention be given to the precise descriptive picture of the borderline histology being analyzed when assessing its significance.

Most studies have found that isolated foci of tubulitis without substantial interstitial inflammation is associated with little or no clinical pathology [37, 45]. However, there is some disagreement about this in the literature. Mehta et al. found that tubulitis in the absence of significant interstitial infiltrate (i.e., at least ≥10% of the cortical area) is associated with a subsequent increase in the incidence of DSA and IFTA, although the incidence of subsequent AR is substantially less that what is associated with t ≥ 1 i ≥ 1 [45]. Mehta et al. also found that the presence of interstitial inflammation in the absence of tubulitis demonstrated non-pathological clinical outcomes similar to what was seen with biopsies having no inflammation [45, 46].

In the context of SCR, the importance of borderline histology is underscored by the fact that 60%–80% of SCR is attributed to borderline changes. Moreover, a number of studies have found that borderline changes are a form of alloimmune risk [2, 29, 47, 48]. In this context, borderline changes have been associated with decreased eGFR, persistent histological inflammation, tubulitis, subsequent AR, de novo DSA, IFTA, and graft loss [37]. This relationship has been observed in surveillance as well as in indication biopsies [8, 39, 45].

Even so, there continues to be a difference of opinion about whether borderline changes invariably represent a consistent rejection phenotype [12, 35, 47]. Evidence has been presented suggesting that borderline changes may be heterogeneous and might not have the same pathological consequences in all cases [37]. Nankivell found that approximately 60% of those with borderline findings in surveillance biopsies and no treatment had subsequent indolent courses while this was not the case with indication biopsies. In patients with borderline changes in surveillance biopsies who underwent rebiopsy, 72% of subsequent biopsies improved, 19% persisted, and 9% worsened. Given these findings, the authors concluded that overall, borderline change suggestive of TCMR is a heterogeneous diagnostic finding that ranges from mild inconsequential inflammation to clinically significant TCMR capable of mediating immune tubular injury with resulting functional, immunological, and histological consequences [37].

Heterogeneity is also suggested by findings of recent detailed but preliminary studies describing certain noninvasive biomarkers which identify “high risk” borderline changes associated with adverse graft outcomes, as distinct from those conferring low clinical risk. These biomarkers include, but are likely not limited to, peripheral blood gene expression profiles and transitional-1 B cells with certain patterns of cytokine expression [49, 50]. Such studies need validation in much larger cohorts. But if validated, studies establishing risk stratification can go a long way to resolving the significance of borderline changes [51].

4.2 Should SCR and Borderline Changes Be Treated?

This is a nuanced topic. In 2009, KDIGO Guidelines suggested treating clinical AR and SCR including borderline changes [52]. However, this was a “suggestion” rather than a “guideline recommendation”. Moreover, the quality of evidence was low and conflicting. KIDIGO guidelines identified seminal studies showing that treatment of SCR improved graft outcomes; but other studies found no benefit from treatment [53-55]. The status of treatment for borderline changes was even more uncertain [52], possibly due at least in part to the changing definition. Other contemporaneous critiques agreed with the 2009 KDIGO treatment guidelines [56].

Most recent studies have demonstrated that treatment of SCR with the phenotypic characteristics of TCMR improves clinical outcomes. Early diagnosis and treatment improves kidney function, the incidence of repeat TCMR, allograft fibrosis, and long-term graft survival [6, 8, 29, 30, 36, 39, 57, 58].

The question of treatment for borderline changes is less clear. There is a robust literature supporting the view that borderline changes defined by the 1997/2019 Banff criteria (t1, i1) represent some form of TCMR less developed than TCMR Banff 1A (defined by t ≥ 2, i ≥ 2). These borderline findings appear to be associated with damaging consequences, especially when they were not treated [8, 39, 45]. Overall, kidneys with treated borderline changes show significantly higher recovery of renal function and improved graft survival [8, 35, 37].

But the evidence for this is not clearcut. As noted above, some have observed that treatment of borderline changes is not uniformly necessary when found in surveillance biopsies. On the other hand, other studies have demonstrated the need for treatment when borderline changes are found in surveillance biopsies [8, 35, 37, 39, 45].

When the determination is that they merit treatment, borderline changes are often treated with high-dose steroids. When borderline changes show only modest interstitial inflammation, many centers prefer to simply augment baseline immunosuppression [37, 45, 60].

4.3 What Is the Status of Surveillance Biopsies to Diagnose SCR in Pediatric Kidney Transplantation?

Currently, SCR is diagnosed by surveillance biopsy, with the rationale that early diagnosis and treatment is critical to improve allograft survival [28, 29, 38].

Most pediatric transplant centers do not perform surveillance biopsies [29, 36, 61]. A recent report found that of 67 pediatric nephrologists who answered a survey, only 34% (23 of 67) performed surveillance biopsies [61]. We also performed a survey that yielded similar results (Table 1). This is also true in adult programs; less than half of large adult transplant centers perform surveillance biopsies [62].

Surveillance biopsies have several practical difficulties, particularly in children. These tend to dissuade many centers from performing them. Biopsies may be technically difficult [6, 36]. Complications such as bleeding and arteriovenous fistulae, while uncommon, do occur [6, 36, 61, 63]. Pathology may be missed due to sampling error [61, 64, 65]. Additionally, surveillance biopsies can be expensive [36]. Finally, there is a practical limit to the number of biopsies that can be performed in any one patient [47, 66, 67].

There is no consensus about the best time to perform surveillance biopsies [68]. Most centers perform intermittent surveillance biopsies in the first 2 years; this is a potential weakness, as there are numerous late AR episodes in pediatric kidney transplant recipients and these have potentially serious consequences for subsequent graft failure [69].

Thus, the cost/benefit ratio as well as the optimal scheduling of surveillance biopsies remains open for debate [70]. Clearly, there is an unmet need for less invasive ways to identify SCR [71].

4.4 SCR and Medication Nonadherence: A Particularly Vexing Problem in Adolescents

Adolescents with ESKD, along with young adults, have the highest rate of allograft loss of any age group [72]. This allograft loss is thought to be due in part to immunosuppressive medication nonadherence [73]. Estimates of the frequency of adolescent nonadherence vary, ranging from 30% to 50% [5, 74, 75]. Medication nonadherence often leads to both AR and de novo DSA and ultimately graft loss [5, 76, 77]. Early on, rejection activity may be clinically unrecognizable, that is, SCR [4]. This under-immunosuppression/SCR smolders until the damage becomes clinically evident. This in turn leads to long-term allograft damage [5, 14].

Nonadherence is notoriously difficult to detect and treat, related as it is to AMR [76, 78-81]. This is often why centers elect to perform surveillance biopsies [61]. A reliable noninvasive biomarker that would obviate the need for surveillance biopsies is a major unmet need in this context.

4.5 Biomarkers for SCR in Pediatric Kidney Transplantation

A biomarker is defined as “a characteristic that is objectively measured and evaluated as an indicator of normal biological processes, pathogenic processes or pharmacological responses to a therapeutic intervention.” [82, 83]. In kidney transplantation, biomarkers are designed to detect, among other conditions, SCR and clinical AR, infection and prediction of and recovery after AR treatment. The performance characteristics of a biomarker are established by comparing test results to kidney histology in specific clinical trials [40]. The utility of a biomarker is established in a training cohort study but must be validated in a rigorous independent study cohort to establish the biomarker's clinical value.

Validated biomarkers may be able to dramatically improve the care of pediatric kidney transplant recipients [40, 84]. A robust biomarker could pre-empt the need for most surveillance biopsies, allow early intervention to reverse SCR and monitor the success of treatment [28, 33, 47, 71, 85]. Key attributes of an ideal biomarker are outlined in Table 2.

Biomarker performance is statistically evaluated using the metrics of specificity and sensitivity, both separately and in receiving operating characteristic (ROC) curve analysis, specifically the area under the ROC curve (AUC). The higher the AUC (with a scale that ranges from 0.5 to 1.0) the greater the biomarker's statistical reliability [86]. Additionally, the clinical value of the biomarker is evaluated by the calculation of the negative (NPV) and positive predictive values (PPV) [22, 85]. These values are established in the training study but must be confirmed in one or more validation studies in order to establish the ultimate clinical utility of the biomarker in question.

Both the PPV and the NPV have important applications [83]. The PPV yields the percentage of patients testing positive who actually have rejection. The NPV yields the percentage of individuals who do not have rejection if the test is negative. That is, the higher the test's NPV, the more likely that a negative result correctly denotes no rejection [83]. A test with a high NPV has a low false negative rate and minimizes missed AR/SCR [87]. Thus, an assay with high utility for avoiding unnecessary surveillance biopsies is characterized by a high NPV. On the other hand, an assay that would provide justification for a for-cause biopsy should have a low false-positive rate and therefore a high PPV [84, 88].

In the context of validating transplant biomarkers, it is important to take note of the prevalence of the diagnosis of rejection in the study cohort. PPV and NPV are affected by the prevalence of the outcome variable, in this case AR. Lower prevalence rates tend to increase the NPV while higher prevalence tends to increase the PPV. As the prevalence of rejection in most studies is relatively modest, this tends to increase the NPV [83]. However, it should be noted that this has sometimes led to over-optimistic assessments of a biomarker's utility.

The pediatric recipient is best served by a biomarker that meets a series of needs: (1) the ability to identify rejection that is not yet accompanied by increasing creatinine [13]; (2) the identification of stable patients without rejection, obviating the need for a surveillance biopsies; (3) identifying stable patients with increasing serum creatinine due to non-rejection causes (e.g., increasing muscle mass with maturation, dehydration, and elevated calcineurin levels); (4) the capacity to detect SCR in recipients with possible medication nonadherence or inadequate immunosuppression; and (5) the detection of successful treatment of AR. Each of these needs is important, and not all biomarkers will address all of these needs.

Only a small number of reports have examined the status of promising biomarkers in pediatric kidney transplantation [40, 89, 90]. In point of fact, a recent exhaustive examination of kidney transplant biomarkers, reported that there were only 37 (5%) pediatric studies identified out of 676 total biomarker studies published [91]. Nevertheless, the field in pediatrics is rapidly maturing. Some promising biomarkers in adult and pediatric kidney transplantation are discussed below. This discussion is limited to assays in either blood or urine. Molecular markers for rejection obtained from biopsy tissue are beyond the scope of this discussion. However, well-validated identification of relevant molecular transcripts in biopsy tissue may soon replace standard histology as the “gold standard” for the identification of AR [92]. Importantly, all of the major molecular pathology studies have excluded children <18 years of age, so it is unknown whether these molecular phenotypes are relevant in pediatric kidney transplant recipients.

We have limited our discussion to those techniques that have been published, being investigated, or have promise in pediatric kidney transplant recipients. Performance characteristics are listed when available. However, because of design and methodological variability and inconsistencies, it is very difficult to compare the quality of current biomarker studies [91]. This is not meant to be an exhaustive list since there are many other techniques in the pipeline.

5.1 Anti-HLA Donor-Specific Antibody (DSA)

Serial monitoring of DSA is perhaps the most widely used immunological posttransplant biomarker. Such monitoring is used to identify the persistence of pretransplant antibodies as well as to discover the generation of de novo DSA. Patients with persistence of preformed DSA have a high risk of AMR and premature graft loss. Development of de novo DSA has been linked with both subclinical and clinical AMR and allograft loss [93-97]. This has been demonstrated in both children and adults (AUC = 0.73) [98-100]. The development of de novo DSA is associated with decreased 10-year graft survival [99]. The detrimental effects of de novo DSA are particularly evident in nonadherent pediatric recipients [101, 102]. Persistence of de novo DSA following treatment of AMR is associated with progressive allograft dysfunction, microvascular inflammation, and transplant glomerulopathy [97].

Current guidelines suggest monitoring for DSA when immunosuppression is being reduced, when there is concern about medication nonadherence, or when there is a rejection episode [103, 104]. Consensus guidelines also recommend routine DSA screening [104]. In a survey of 30 pediatric centers, 86% performed routine surveillance for de novo DSA. The predominant schedule entailed screening every 3 months in the first year after transplant, and every 6 months in the second year [105]. Advance molecular matching and/or eplet matching may help ameliorate the generation of de novo DSA [5, 106].

5.2 Non-HLA Antibodies

In selected patients with AMR, particularly those that are not accompanied by de novo DSA, pathogenic agonistic antibodies directed against non-HLA antigens have been identified [107-110]. The most completely characterized are those directed against angiotensin II Type 1 receptors (AT1R); however, others have been identified [3, 111]. These antibodies are strongly associated with decreased eGFR, significant hypertension and histological findings of AMR. The histological/molecular characteristics of the AMR seen with AT1R antibody are characterized as a complement-independent vascular rejection with intense vascular inflammation, and inflammatory cytokine deposition. There is an absence of c4d capillary deposition [111-113]. These antibodies, and their attendant decrease in eGFR, are more common in pediatric patients than in adults [3, 111, 112]. AMR in the absence of DSA should prompt a search for such non-HLA antibodies.

5.3 Immunosuppressive Drug Levels

Variability in tacrolimus and/or sirolimus drug levels is associated with DSA and biopsy-proven rejection (BPAR) [76, 79]. The variability can be statistically assessed in a number of ways including using the percent coefficient of variation. Increased variability has been associated with medication nonadherence. There are published data in both pediatric and adult kidney transplant recipients [76, 79, 114].

6.1 Gene Expression Panels in Peripheral Blood

These tests appear to identify SCR, either with borderline changes or TCMR. While they appear to be promising in individual study cohorts, there are, as yet questions in independent datasets [115]. In general, the tests are often trained against a series of biopsies consisting of both for- cause and surveillance biopsies. Importantly, there are not yet published gene expression panel studies that have been performed using only surveillance biopsies in children to identify SCR.

7.1 Three gene panel of CD3ϵ mRNA, IP10 mRNA, and 18S rRNA [126]

In a study in adults, this assay of urinary cells identified TCMR before its clinical appearance with an AUC = 0.74–0.83, a sensitivity = 71% and specificity = 72%. The test may be confounded by BK viral infection [90]. The original test presented logistical challenges in specimen preparation [127] but newer technique modifications have successfully addressed this issue. This panel is currently undergoing pediatric validation in the VIRTUUS trial [ClinicalTrials.gov: NCT03719339] [128].

7.2 Other Urinary Gene Expression Profiles

Based upon genes identified in rejecting kidney allograft tissue, Sigdel et al. identified an 11 gene urinary cell panel strongly associated with AR. It did not differentiate TCMR from AMR. The study population had a substantial proportion of pediatric recipients, but discrete pediatric results were not reported [129]. Verma et al. used urinary cell transcriptomic profiling/RNA Sequencing (RNA Seq) to identify both TCMR and AMR [130]. El Fekih et al. isolated mRNA from urinary exosomes and developed a signature with excellent performance characteristics for AR [131]. In a recent study including children reported by Hirada et al. RNA markers from urinary exosomes and macrovesicles identified TCMR and allograft injury with a high degree of significance [132]. Additionally, a urine assay for FOXP3 mRNA performed well in predicting acute clinical rejection (AUC = 0.76; 90% sensitivity; 73% specificity) [133].

8.1 Urinary Chemokines

Urinary IFNγ-dependant chemokines (measured as either protein or mRNA), indicate kidney transplant injury [134]. These include urinary CXCL9 and CXCL10, biomarkers of graft inflammation [135]. Multiple studies have established excellent performance as markers for AR [136, 137]. In a study of urinary CXCL9 and CXCL10 in adults and children, both chemokines were elevated in AR; however, both were also elevated in BKV infection [121]. CXCL9 and CXCL10 show similar performance characteristics [138]. It has been suggested that these urinary chemokines may perform best when combined with other biomarkers and/or clinical parameters [139, 140].

8.2 Urinary CXCL9 mRNA [AUC = 0.79] and Protein [AUC = 0.86] [141, 142]

Both urinary CXCL9 mRNA and protein are strongly correlated with interstitial inflammation and tubulitis and can distinguish subclinical tubulitis from normal or borderline histology [141]. CXCL9 protein has a high NPV (92%) and a moderate PPV (68%) and may be elevated for up to 30 days prior to clinical BPAR [141]. Similar performance statistics hold for CXCL9 mRNA. Levels of CXCL9 protein can also diagnose SCR (AUC = 0.78) possibly for both TCMR and AMR [137]. Low levels of CXCL9 protein associated with low rates of subsequent rejection [143, 144]. A rapid, inexpensive point-of-care assay, using CRISPR-Cas 13 technology, can identify urinary CXCL9 mRNA allowing for home or point-of-care monitoring [145].

8.3 Urinary CXCL10 mRNA and Protein

Most studies examining CXCL10, and particularly those in pediatrics, utilize CXCL10 protein normalized to urine creatinine concentration. This assay can detect SCR (AUC = 0.81), clinical TCMR (AUC = 0.88), and AMR [146]. There is a robust pediatric literature [89, 147]. In a recent multicenter pediatric trial, urinary CXCL10 was significantly associated with BPAR (AUC = 0.73–0.76). This assay also tracks recovery from rejection after successful treatment. CXCL10 protein can be elevated up to 4 weeks prior to clinical AR [147]. CXCL10 protein demonstrates a high NPV for AR. A low level of CXCL 10 suggests immune quiescence. When used with de novo DSA, the diagnostic performance for AMR is enhanced (AUC = 0.83) [148]. Similar to CXCL9, point-of-care technology (see above), is also available for CXCL10 mRNA [145, 149]. A recent large randomized controlled trial in adults examined the clinical utility of urinary CXCL10. While this assay was useful in identifying SCR, the trial failed to find any clinical benefit at 1-year posttransplant [150]; it is likely that 1-year clinical parameters may not be the best outcome to use. Longer-term follow-up is needed to fully establish the value and clinical utility of this assay [151, 152].

8.4 Urine Metabolomics

The NIH CTOT-04 study identified a series of promising urinary metabolites in cell-free supernatants that identified TCMR with AUCs of 0.75–0.81. Moreover, in this same study, when urinary measurements for 18SrRNA, CD3Ɛ mRNA, and IP-10 mRNA (see above) were combined with the metabolite values, the AUC increased to 0.91 [103, 153]. In another study, spectroscopy identified 134 urine metabolites indicating borderline and/or TCMR (AUC = 0.90). Findings were similar in both pediatric and adult cohorts [154-156]. Recently, a different metabolite set, with even greater sensitivity and specificity for BPAR in children has been described [157]. Either by themselves or by used in conjunction with other assays, the identification of target urine metabolomics may hold great promise. However, limitations of a metabolomics approach include the paucity of appropriate clinical laboratory resources [158].

9.1 Allosure (CareDx)

This test has the longest clinical track record. The test analyzes 405 SNPs [181]. Adult data suggest that performance is best for AMR (AUC = 0.82–0.87). The pivotal DART study (Diagnosing Acute Rejection in Kidney Transplant) found that the level of %dd-cfDNA in patients with AMR ranged from 1.4% to 2.9%, while for TCMR, the results averaged approximately 1.2% [11, 182]. Some data suggest that TCMR (Banff 1A and/or 1B) and borderline changes may be associated with a threshold of 0.5% [12, 183]. However, more data are needed [184].

Published clinical experience with Allosure in pediatric kidney transplant recipients has been increasing. A recently published study in pediatric recipients undergoing for-cause biopsies for the suspicion of rejection (N = 48) showed Allosure %dd-cfDNA >1% was significantly associated with both DSA and BPAR (sensitivity = 86%; specificity =100%; AUC = 0.996) [160]. As with adult studies, the test performed best identifying AMR. The ability to document resolution of rejection after treatment was unclear; however, the patient sample was small [167]. Using a longitudinally followed cohort, another pediatric study using Allosure also confirmed the level of ≥1% discriminates BPAR from quiescence with an AUC of 0.82 [168].

9.2 Prospera (Natera)

This dd-cfDNA test analyzes over 1300 SNPs utilizing a massively multiplex PCR technique [185]. In a study of 277 plasma samples from 178 patients, 38 samples were obtained from patients with BPAR. At a cutoff of 1%, the AUC was 0.87, the sensitivity was 89%, and the specificity was 73%. Based on a 25% prevalence of rejection, the PPV was 52%, and the NPV was 95% [163, 185]. Both AMR and TCMR had the same median levels of %dd-cfDNA, suggesting that elevated Prospera %dd-cfDNA can detect TCMR [92, 185]. While 20% of patients in the total cohort were <18 years none had BPAR and thus no conclusions can as yet be drawn about its use in pediatrics [185].

9.3 TRAC [Transplant Rejection Allograft Check] (Viracor Eurofins)

This platform uses an advanced next-generation sequencing technique. Performance characteristics are similar to the other two platforms [163]. The sensitivity/specificity is virtually identical to Allosure, while the AUC is very close to that of Prospera. Early data suggest a cutoff of 0.7%. No pediatric data have yet been reported.

10.1 Torque Teno Virus (TTV)

TTV is a prominent member of the human virome as part of the Anelloviridiae virus family. It is a benign virus that has no human disease associations. It has a potentially promising role as a transplant biomarker for a number of reasons: (1) In most populations, there is a universal presence of TTV viremia; (2) TTV viremia appears to mirror the global level of immunosuppression in solid organ transplant recipients—the higher the TTV copy number the greater the degree of immunosuppression; and the lower the viral load, the lower the degree of immunosuppression; (3) a series of recent studies and meta-analyses suggest that a high TTV viral load is associated with posttransplant infection, while a low viral load has been strongly associated with rejection activity [186-189]. Of note, the assessment of low TTV viral load has been associated with SCR [190]. One variable that must be defined are the kinetics of TTV as it relates to immunosuppressive treatment—namely, how long after a change in immunosuppressive treatment will the viral load reach equilibrium [191]. This has begun to be addressed in adults, but there are no data in children.

Commercially standardized testing has been developed. Reports suggest that various techniques display good testing characteristics with comparable results, albeit with some discrepancies [192].

Currently, data in pediatric kidney transplant recipients are limited. Uhl et al studied 45 pediatric kidney transplant recipients longitudinally for a period of 1 year starting at least 3 months after transplantation. As with adult reports, this study found that TTV DNAemia was associated with the strength of immunosuppression, as assessed by the strength of certain immunosuppressive medications [193]. A more recent pediatric study conducted by Eldar-Yedida et al. found a significant association between rejection episodes and low TTV viral load [194], similar to what has been reported in adult kidney transplant recipients.

The identification of a patient's global immune function could be an important step forward in the quest to personalize posttransplant immunosuppression. A randomized controlled interventional blinded Phase II trial in adults has been initiated [186]. Patients will be managed either in the standard fashion guided by tacrolimus blood levels or by dosing immunosuppressive agents based on TTV viral loads.

10.2 ImmunoKnow (Cylex)

This is a blood test measuring adenosine triphosphate (ATP) from CD4+ T-cell responsiveness to a mitogen to measure the level of transplant recipient immunosuppression. Initial studies, which included pediatric transplant recipients, found that results were correlated with immune status and AR [195, 196]. However, a number of other studies were unable to validate this test's utility [197-199].

10.3 Other Novel Techniques and Biomarkers

There have been several other studies published examining novel molecules and techniques that are candidate biomarkers for rejection. In this regard, there has been a good deal of interest in the isolation and interrogation of extracellular vesicles in both peripheral blood and urine. Urinary and plasma exosomes/vesicles have been utilized in several studies examining molecules of interest including various proteins, T cell–derived vesicles, and specific genes [200]. These are particularly useful in isolating immunologically important molecules in the rejection process [200]. Their use in the diagnosis of rejection has been reported but they are not yet established clinically.