Urinary Expression of Kidney Injury Markers in Renal Transplant Recipients

Abstract

Background and objectives: The outcome of renal transplantation after an episode of acute rejection is difficult to predict, even with an allograft biopsy. We examined whether urinary expression of specific biomarker mRNA could be used as a noninvasive prognostic marker in kidney transplant recipients.

Design, setting, participants, & measurements: We studied 63 kidney transplant recipients who require graft biopsy because of progressive worsening of kidney function. The mRNA of neutrophil gelatinase-associated lipocalin, kidney injury molecule-1 (KIM-1), IL-18, surfactant protein-C, and S100 calcium-binding proteins A8 and A9 in urinary sediment were quantified.

Results: Urinary expressions of neutrophil gelatinase-associated lipocalin, KIM-1, and IL-18, but not other target genes, were significantly different between histologic groups (P < 0.0001 for all). After followed for an average of 39.7 ± 21.1 months, the rate of renal function decline significantly correlated with urinary KIM-1 expression (r = −0.434, P = 0.0004) but not other target genes. At 48 months, the graft survival rate for the high and low KIM-1 groups were 46.2 and 78.6%, respectively. After adjusting for confounding variables, each log of higher urinary KIM-1 expression conferred an ∼2.9-fold higher risk of developing graft failure (95% confidence interval, 1.3- to 6.2-fold; P = 0.006). The result remained similar when only patients with no acute cellular rejection were analyzed.

Conclusions: In kidney allograft recipients, urinary KIM-1 expression provides prognostic information in relation to the rate of renal function decline, irrespective of the kidney pathology.

Chronic progressive loss of kidney allograft function, previously known as chronic allograft nephropathy (CAN), is the major cause of late graft failure in kidney transplant recipients (1). The number, timing, and severity of acute rejection episodes are the strongest predictors of CAN and graft failure (2–4). However, not all acute rejection episodes lead to CAN. For example, the occurrence of acute vascular rejection is associated with a more adverse prognosis than the occurrence of acute cellular rejection (5). The nature of the inflammatory infiltrate and the restorative effect of the renal tissue on acute rejection may also partly account for patients' predisposition to develop CAN (6), whereas changes in mRNA levels of inflammatory components and extracellular matrix–regulating molecules within the first 12 months after transplantation are associated with progressive allograft dysfunction (7–10).

There has been much effort in identifying noninvasive prognostic markers for kidney transplant recipients and nontransplant patients with acute kidney injury. To date, a number of urinary biomarkers have been developed for the prediction, diagnosis, and risk stratification for patients presenting with acute kidney injury in general (11–13). Based on available literature, the following candidates have the best potential of further development for clinical application (14): neutrophil gelatinase-associated lipocalin (NGAL), kidney injury molecule-1 (KIM-1), and IL-18. For patients who newly received a kidney transplant from deceased donors, a recent prospective study showed that urinary NGAL and IL-18 are early noninvasive predictors of both the need for dialysis within the first week of transplantation and 3-month recovery of graft function (15). However, the role of these biomarkers in the setting of the kidney transplant recipient has not been explored.

As to kidney transplant, various reports have emphasized the capacity of microarray technology to find genes and elucidate molecular pathways that are involved in the progression of renal allograft damage (16–21). Identification of prognostic factors in the transplantation setting may lead to the development of improved intervention strategies and may contribute to a better understanding of the pathophysiology of CAN. Recently, Sarwal et al. (10) identified three distinct subtypes of acute rejection with cDNA microarray profiling and showed that the composition of the infiltrate is related to the prognosis, and Eikmans et al. (22) further confirmed that intrarenal mRNA expression of surfactant protein-C (SP-C), S100 calcium-binding proteins A8 (S100A8) and A9 (S100A9) were associated with the prognosis of kidney transplant recipients.

The clinical application of quantifying intrarenal gene expression, however, is hampered by the need for renal biopsy. In the last few years, with the development of reliable RNA extraction techniques from urinary sediment and reverse transcription real-time quantitative PCR, measurement of mRNA expression in urinary sediment has become an emerging tool for the study of kidney diseases (23). In this study, we aim to examine whether urinary expression of the above mentioned target genes could be used as a noninvasive prognostic marker in kidney transplant recipients.

Materials and Methods

Patient Selection

We recruited 63 kidney transplant recipients who require graft kidney biopsy because of suspected acute rejection or progressive worsening of kidney function. Patients with obvious urinary tract infection or acute tubular necrosis as the cause of acute renal function deterioration were excluded. The study was approved by the ethics committee of the Chinese University of Hong Kong. All patients had written informed consent. On the day of kidney biopsy, a whole-stream early morning urine specimen was collected for mRNA extraction and determination of gene expression. We also studied 11 stable kidney transplant recipients as controls; whole-stream early morning urine specimen was collected during routine follow-up.

Preparation of Urinary Sediment

The methods of urinary sediment isolation and mRNA extraction have been described previously (24). Briefly, urine samples were centrifuged at 3500 × g for 30 minutes at 4°C. Total RNA in urine and blood samples was extracted by the RNeasy Mini Kit and QIAamp RNA Blood Mini Kit (Qiagen), respectively. For each reaction, ∼0.5 μg of RNA was reverse transcribed to cDNA with the Superscript II RNase H-Reverse Transcriptase (Invitrogen).

Quantification of mRNA Expression

We used the real-time quantitative PCR by the ABI Prism 7700 Sequence Detector System (Applied Biosystems, Foster City, CA) to quantify the mRNA expression from urinary sediment. As suggested by previous studies on acute kidney injury, we quantified the mRNA expression of NGAL, KIM-1, and IL-18 (13). Based on the study of Eikmans et al. (22), we also quantified SP-C, S100A8, and S100A9 expression. Taqman primers and probes of each target were purchased from Applied Biosystems. The mRNA expression for each signal was calculated by using the ΔCt procedure according to manufacturer's instruction. The amount of target mRNA was controlled by glyceraldehyde 3-phosphate dehydrogenase as the housekeeping gene and expressed as number of mRNA copy per 100,000 copies of the housekeeping gene.

Morphometric Study of Kidney Biopsy

After routine histologic and immunofluorescence study, we performed morphometric analysis of the kidney biopsy specimen to quantify the degree of renal scarring. Jones' silver staining was performed on 5-μm-thick sections of renal biopsy specimen of each patient. As described previously (25), we used computerized image analysis method to semiquantify nephrosclerosis. Ten randomly selected areas were assessed in each patient, and the average percentage of scarred tubulointerstitial areas, as represented by the percentage of the area with positive staining, was computed for each patient.

Clinical Management

All renal biopsy specimens were assessed by a single pathologist (F.M.M.L.) and diagnosed according to the Banff 2007 criteria (26). Histologic rejection is defined by a subsequent biopsy performed within 30 days with Banff acute score indices i2t2v0 or above (26). Clinical management of the individual patient was decided by the individual nephrologist and was not affected by the study. In general, acute cellular rejection was treated with pulse intravenous methylprednisolone. Complete response was defined as a serum creatinine level <110% of the baseline value at 40 days after the initial biopsy (27). Partial response was defined as serum creatinine level ≥110% of the baseline value and <90% of the peak value. Patients who had serum creatinine >90% of the peak value after 40 days or those required intravenous anti-thymocyte gamma globulin were considered as no response. Nonspecific tubular atrophy and interstitial fibrosis (TA/IF) were generally managed by minimization of calcineurin inhibitor dose, tight control of BP, and addition of angiotensin-converting enzyme inhibitor.

Clinical Follow-up

After renal biopsy, all stabilized patients were followed at least every 8 weeks for at least 12 months. During each follow-up, serum creatinine and spot urine protein-to-creatinine ratio were measured. GFR was estimated by the Nankivell formula (28). Primary outcome measures were the rate of GFR decline and graft survival. The rate of GFR decline was calculated by the least-square regression method. Follow-up for graft survival was administratively censored on September 30, 2009.

Statistical Analyses

Statistical analyses were performed by SPSS for Windows software version 15.0 (SPSS, Chicago, IL). Descriptive data are represented as mean ± SD. For the simplicity of analysis, histologic diagnosis was classified into three predefined groups: rejection; nonspecific tubular atrophy and interstitial diagnosis; and other specific diagnosis (for example, recurrence of original disease). Because the gene expression data were highly skewed, the gene expression data were compared by the Kruskall-Wallis test and Spearman's rank correlation as appropriate. P < 0.05 was considered significant. All probabilities are two-tailed.

For the analysis of graft survival, urinary gene expression data were log-transformed before fitting into the survival model. Because univariate analysis suggested that urinary KIM-1 expression correlated with graft survival, we constructed a Cox proportional hazard model for the analysis of graft survival. In addition to using urinary KIM-1 expression, we included patient age, duration after transplantation, histologic diagnosis of graft biopsy, baseline GFR, proteinuria, and degree of histologic scarring by morphometric study as independent variables. These parameters were selected for the construction of the Cox model because they were generally accepted as important predictors of graft survival.

Results

We recruited 63 kidney transplant recipients (56 from deceased donors; 7 living-related donors) and 11 controls (all except one from deceased donors). Kidney allograft biopsy showed acute cellular rejection (25 cases), nonspecific TA/IF (25 cases), and other specific pathologic entities (13 cases). Their baseline clinical characteristics are summarized in Table 1. In general, the control group had better renal function, lower proteinuria, and received a lower dose of immunosuppressive therapy than the other groups. However, there was no significant different in any baseline clinical parameter between the three study groups.

Table 1.

Baseline demographic and clinical characteristics

Group	TA/IF	Rejection	Others	Control
No. of patients	25	25	13	11
Gender (M:F)	12:13	16:9	11:2	7:4
Age (years)	45.9 ± 12.7	45.1 ± 10.5	45.5 ± 8.0	45.2 ± 11.0
Living-related transplant, no. of cases (%)	3	2	2	1
Duration after transplant (months)	98.4 ± 69.7	61.3 ± 65.0	75.7 ± 54.2	80.0 ± 67.9
Diagnosis, no. of cases (%)
IgA nephropathy	6	11	5	4
Other glomerulonephritis	10	8	5	3
Diabetic nephropathy	0	1	1	0
Hypertension	1	0	0	0
Obstruction	2	2	0	1
Others/unknown	6	3	2	3
Immunosuppressive regimen, no. of cases (%)
Steroid + cyclosporin	10	4	4	2
Steroid + azathioprine + cyclosporin	13	15	7	8
Steroid + mycophenolate + cyclosporin	1	4	1	1
Steroid + azathioprine + tacrolimus	1	2	1	0
Clinical parameters
Serum creatinine (μmol/L)	265.6 ± 116.3	295.7 ± 228.1	214.5 ± 111.4	178.0 ± 38.4
Estimated GFR (ml/min per 1.73 m2)	25.2 ± 13.6	24.7 ± 11.2	30.6 ± 12.7	36.5 ± 10.0
Proteinuria (g/day)	0.94 ± 0.60	0.89 ± 0.29	0.73 ± 0.23	0.07 ± 0.14
Drug treatment
Prednisolone (mg/day )	7.9 ± 5.7	10.2 ± 9.9	5.5 ± 3.1	4.8 ± 1.4
Cyclosporin (mg/day)	177.3 ± 61.7	187.5 ± 75.1	160.4 ± 37.8	133.2 ± 39.5
2-h cyclosporin level (μg/L)	944.0 ± 599.9	891.0 ± 288.0	728.2 ± 229.9	656.0 ± 108.7

TA/IF, nonspecific tubular atrophy and interstitial fibrosis.

Relation with Baseline Characteristics

Urinary expression of NGAL, KIM-1, IL-18, SP-C, S100A8, and S100A9 did not correlate with baseline clinical parameters, including patient age, duration of transplantation, estimated GFR, proteinuria, and dose of prednisolone or cyclosporine (data not shown). Urinary expression did not correlate with the degree of TA/IF scarring in any of the target genes (data not shown).

Urinary expression of the target genes for specific histologic diagnosis groups is shown in Figure 1 and Table 2. Urinary expression of NGAL, KIM-1, IL-18, S100A8, and SP-C was different between groups (Kruskal-Wallis test: P < 0.0001, P < 0.0001, P < 0.0001, P = 0.049, and P = 0.07, respectively), although the result for SP-C did not reach statistical significance. In contrast, urinary expression of S100A9 was not significantly different between diagnosis groups (data not shown). Further subgroup analysis showed that urinary expression of NGAL, KIM-1, and IL-18 (but not that of SP-C or S100A8) was significantly different between histologic diagnosis groups (data not shown).

Figure 1.

Comparison of gene expression in urinary sediment between diagnosis groups. Data are compared by Kruskal-Wallis test. Gene expressions are expressed as number of copy per 100,000 copies of the housekeeping gene. The box indicates median and 25 and 75 percentiles; whisker caps indicate 5 and 95 percentiles; and outliers are represented as closed circles (key for diagnosis groups: TA/IF, nonspecific tubular atrophy and interstitial fibrosis; reject, acute cellular rejection; other, other specific pathology; CTL, control with stable renal function).

Table 2.

Gene expression in urinary sediment between diagnosis groups

Group	TA/IF	Rejection	Others	Control
No. of patients	25	25	13	11
NGAL	69.5 (34.3 to 382.0)a	10.4 (16.5 to 39.8)b,c	17.2 (6.9 to 25.6)	3.3 (6.8 to 17.2)
KIM-1	25.4 (12.9 to 109.5)a	3.8 (1.4 to 16.0)a,d	3.8 (3.8 to 15.4)a	0.2 (0.1 to 0.9)
IL-18	365.9 (177.3 to 798.2)a	71.7 (37.8 to 147.2)b,d	80.3 (17.1 to 146.3)	0.9 (0.4 to 35.7)
SP-C	6.1 (0.6 to 27.0)	0.7 (0.2 to 6.6)	2.2 (0.3 to 16.3)	4.4 (2.9 to 31.7)
S100A8	127.4 (46.7 to 630.7)	96.1 (25.9 to 736.3)	123.3 (35.6 to 689.6)	378.0 (378.0 to 1396.6)
S100A9	212.8 (55.4 to 2132.1)	244.2 (58.4 to 1160.5)	148.3 (60.6 to 875.2)	187.7 (40.3 to 2252.1)

Gene expressions are expressed as number of copy per 100,000 copies of the housekeeping gene; data are presented as median (25 to 75 percentile).

TA/IF, non-specific tubular atrophy and interstitial fibrosis.

^aP < 0.001 and

^bP < 0.01 compared with the control group.

^cP < 0.01 and

^dP < 0.001 compared with the TA/IF group.

Response to Rejection Treatment

For the 25 cases with acute cellular rejection, 8 had a complete response, 11 had a partial response, and 6 had no response to anti-rejection therapy. The urinary expression of the target genes is compared and shown in Figure 2. Urinary expression of NGAL and IL-18 was marginally different between treatment response groups (P = 0.004 and P = 0.015, respectively). Urinary expression of KIM-1, SP-C, S100A8, and S100A9 was not significantly different between treatment response groups.

Figure 2.

Comparison of gene expression in urinary sediment between treatment response groups in patients with acute cellular rejection. Data are compared by Kruskal-Wallis test. Gene expressions are expressed as number of copy per 100,000 copies of the housekeeping gene. The box indicates median and 25 and 75 percentiles; whisker caps indicate 5 and 95 percentiles; outliers are represented as closed circles (key for response groups: NR, no response; PR, partial response; CR, complete response).

Relationship with Renal Function Decline

The patients were followed for an average of 39.7 ± 21.1 months; the rate of GFR decline was −4.15 ± 6.49 ml/min per year. The relationship between the rate of GFR decline and urinary gene expression is summarized in Figure 3. In short, the rate of GFR decline significantly correlated with urinary KIM-1 expression (r = −0.434, P = 0.0004) but not other target genes.

Figure 3.

Relation between urinary gene expression and rate of GFR decline. Correlation coefficients are calculated by the Spearman's test. Gene expressions are expressed as number of copy per 100,000 copies of the housekeeping gene.

During the follow-up period, 22 patients progressed to end-stage renal failure and were put on dialysis; 3 patients died of infection with a functioning graft—their graft survival status was censored. With receiver operating characteristic curve analysis, only urinary KIM-1 expression could predict graft failure, with an area under the curve of 0.684 (P = 0.017; Figure 4), whereas urinary expression of other target genes did not have a significant association with graft failure (data not shown). The optimal level of urinary KIM-1 expression for the prediction of graft failure was found to be 15 copies per 100,000 copies of the housekeeping gene. At this level, the sensitivity was 68.2%, specificity was 68.3%, positive predictive value was 65.2%, and negative predictive value was 67.5%. Using this cut-off, the graft survival rate for the high and low KIM-1 groups were 46.2 and 78.6%, respectively, at 48 months (Figure 5). By univariate Cox analysis, urinary KIM-1 expression was significantly associated with graft survival (unadjusted hazard ratio, 3.28; 95% confidence interval, 1.335 to 8.057; P = 0.01).

Figure 4.

Receiver operating characteristic curve of urinary KIM-1 mRNA expression as a predictor of graft failure in 4 years. The calculated area under the curve is 0.684.

Figure 5.

Kaplan-Meier plot of graft survival. Patients are classified into high and low urinary KIM expression. A urinary KIM expression of 15 copies per 100,000 copies of the housekeeping gene (four times above the average expression of healthy control) was taken as the cut-off for high and low KIM expression.

The multivariate Cox model of graft survival is summarized in Table 3. After adjusting for confounding variables, independent predictors of graft survival were duration of transplantation, baseline renal function, proteinuria, and urinary KIM-1 expression. In this model, each log of higher urinary KIM-1 expression conferred an ∼2.9-fold higher risk of developing graft failure (95% confidence interval, 1.3- to 6.2-fold). The result of the multivariate Cox model remained similar when only patients with no acute cellular rejection were included for analysis (data not shown).

Table 3.

Cox proportional hazards models of graft survival

	AHR	95% CI	P
Duration of transplant (years)	1.138	1.036 to 1.250	0.007
Baseline estimated GFR (ml/min per 1.73 m2)	0.903	0.852 to 0.956	0.0005
Baseline proteinuria (g/day)	1.436	1.060 to 1.944	0.02
Urinary KIM expressiona	2.878	1.345 to 6.157	0.006

AHR, adjusted hazard ratio; CI, confidence interval.

^aExpression as log of the number of copy per 100,000 copies of the housekeeping gene.

Discussion

In this study, we found that urinary NGAL and IL-18 expression is different between diagnosis groups in kidney allograft recipients. In patients with acute cellular rejection, urinary expression of these two target genes may also be related to the response with anti-rejection therapy. In contrast, urinary KIM-1 expression provides prognostic information in relation to the rate of renal function decline, irrespective of the kidney pathology.

The use of a gene expression study in renal transplant recipients has been extensively studied (29). Notably, previous studies showed that the mRNA expression of granzyme and perforin in urinary sediment provides a noninvasive means for the diagnosis of kidney allograft rejection (23), and the expression of FOXP3 may predict the outcome of acute rejection in kidney transplant recipients (30). Our result provides further information to support the use of gene expression study in the urinary sediment of kidney transplant recipients.

In this study, we found that mRNA expression of SP-C, S100A8, and S100A9 in urinary sediment does not provide diagnostic or prognostic information for transplant recipients. Our result is in distinct contrast from the previous report by Eikmans et al. (22), who showed that a high expression of S100A8 and S100A9 in the kidney allograft biopsy during acute rejection was associated with a favorable prognosis, whereas high SP-C expression was associated with an unfavorable prognosis. It is important, however, to recognize that Eikmans et al. (22) studied intrarenal gene expression, whereas we quantified the expression in urinary sediment; the two parameters are not necessarily parallel to each other.

Our results on NGAL, KIM-1, and IL-18 share much similarity with studies on acute kidney injury in nontransplant patients (11–13), as well as early after kidney transplantation (15). NGAL is an innate anti-bacterial factor found in activated neutrophils; it is also found in the kidney tubules in response to various injuries (31). IL-18 is produced by activated macrophages and is involved in activation of type 1 T-helper cells. In this study, we found that urinary NGAL and IL-18 expression seemed to be higher in patients with acute rejection or chronic allograft nephropathy (of which immune-related injury contributed to the pathogenesis) than those with other specific pathology (see Figure 1). Unfortunately, although the result was statistically significant, there was a substantial overlap in the urinary expression level between diagnosis groups, and clinical applicability seems unlikely.

On the other hand, our result on KIM-1 as a prognostic marker seems more promising. KIM-1 is a type 1 transmembrane protein expressed on the proximal tubule apical membrane whenever a toxin or pathophysiological state results in dedifferentiation of the epithelium, a phenomenon that possibly precedes irreversible tubular injury (32). In hospitalized nontransplant patients with acute kidney injury, it has been reported that a high urinary KIM-1 level indicates an excess risk of death or need of dialysis (likelihood ratio 1.4) (33), which is in line with our findings. Although Hall et al. (15) found that urinary KIM-1 did not predict recovery of kidney function after transplantation, this study focused on the short-term outcome immediately after kidney transplant. Furthermore, this study quantified urinary KIM-1 at the protein level, whereas the clinical relevance of its urinary mRNA expression has not been explored.

Several limitations of this study need to be addressed. Most importantly, the sample size of 63 was small, and patients had different kinds of renal involvement and heterogeneous treatment protocols. Although the results are statistically significant, many of the differences between groups are modest, and it is not clear whether there is enough separation to distinguish the groups. Furthermore, the control group was also poorly defined; although this group had stable renal function, many of them had a certain degree of renal impairment, and we could not exclude the presence of occult rejection or chronic allograft nephropathy in this group. It remains uncertain whether it is clinically meaningful to combine outcome data for late acute rejection, IF/TA, and recurrent diseases, because they all have different pathophysiologies. There are also numerous confounding factors that could influence response to therapy, and our results should be interpreted with caution. For example, the blood cyclosporin levels were on the high side for patients long after transplantation. Although we did not find overt histologic evidence of calcineurin inhibitor toxicity in our patients, it remains possible that some patients actually had cyclosporin toxicity. In addition, we did not perform serial urine collection or examine the longitudinal trend of gene expression with disease progression. Our previous studies on nontransplant cases, however, showed that urinary mRNA expression has little intraindividual day-to-day variation for stable patients (34,35).

Second, we did not determine the protein levels of the biomarkers in the urine. In theory, determination of the respective proteins has the advantage of an easy and reproductive assay, and it also provides strong additional support for our hypothesis. In contrast, it has also been argued that urinary biomarkers at the protein level could represent leaking from the systemic circulation and correlate with the degree of proteinuria rather than local damage (36). Unfortunately, in this study, we did not save the urinary supernatant for the respective protein level assay.

Furthermore, many correlations were weak. For example, the correlation coefficient of KIM-1 and rate of GFR decline was −0.434, which means that only 18.8% of the variability of GFR decline is explained by KIM-1. We did not perform immunostaining for the urine sediment in this study and, as a result, could not ascertain the cellular origin of mRNA. However, to the best of our knowledge, most of the target genes, except IL-18, are primarily expressed by tubular epithelial cells. Because the microenvironment of urine is quite different from that in renal tissue, the correlation between mRNA expression in the urine sediment and that within the kidney needs further exploration.

Conclusions

In kidney allograft recipients, urinary KIM-1 expression provides prognostic information in relation to the rate of renal function decline, irrespective of the kidney pathology.

Disclosures

None.