Acute rejection of organ allografts remains a formidable problem threatening short- and long-term outcomes (1). The immunosuppressive therapy, used to prevent and treat an episode of acute rejection, is a significant risk factor for life-threatening infectious complications and for the development of posttransplant malignancy (2, 3). Thus, a better understanding of mechanisms of acute rejection and safer, preferably noninvasive, approaches for diagnosing acute rejection and predicting its responsiveness to antirejection therapy are high priorities for the transplantation community.
T lymphocytes are central to the acute rejection process (4) and much has been learned in recent years regarding transmembrane signaling requirements including cell surface proteins contributing to antigen recognition, positive costimulation, and negative regulation (5, 6). There has also been considerable progress in the area of antigen presentation including the discovery of molecules displayed on antigen-presenting cells that function as counter receptors to cell surface moieties on T and B cells (7, 8). Elegant preclinical studies have brought to our attention the critical role of costimulatory receptors and negative regulatory receptors in allograft rejection and tolerance (9–14).
We investigated the gene expression patterns of positive costimulatory molecules OX40 and its ligand OX40L, and the expression profile of negative regulatory proteins programmed death (PD)-1, PD ligand 1 (PD-L1), and PD ligand 2 (PD-L2) in recipients of deceased donor or live donor renal allografts. Our selection of these molecules for mRNA expression profiling was based on preclinical studies demonstrating their participation in allograft rejection and tolerance (9–14) and our reasoning that information regarding their expression patterns in clinical transplantation might be of value from diagnostic, prognostic, and therapeutic perspectives.
Our laboratory has developed a urinary cell mRNA profiling protocol and applied it clinically to ascertain the immune status of the renal allograft (15, 16). In a series of investigations, we identified that mRNA encoding cytotoxic attack molecules granzyme B and perforin are overexpressed in urinary cells during an episode of acute rejection (15) and that the level of mRNA for the regulatory T-cell (Treg)-specification factor Foxp3 in urinary cells is diagnostic and prognostic of acute rejection of human renal allografts (16).
In the current investigation, we tested the diagnostic and prognostic utility of levels of mRNA for OX40, OX40L, PD-1, PD-L1, and PD-L2 in urinary cells. We also investigated whether an algorithm that included our previously identified biomarkers of acute rejection increased the diagnostic accuracy.
We report here that the urinary cell levels of mRNA for OX40, OX40L, and PD-1, but not the levels of PD-L1 or PD-L2, are significantly higher during an episode of acute rejection. A model that included levels of mRNA for OX40, OX40L, PD-1, and Foxp3 was an extremely accurate predictor of acute rejection. Moreover, acute rejection reversibility and graft loss after an episode of acute rejection was predicted by urinary cell levels of mRNA for OX40, OX40L, and Foxp3. The multigene combination resulted in a larger area under the curve (AUC) compared with the AUC of OX40 mRNA alone, and the difference was statistically significant for predicting acute rejection but not acute rejection reversal or graft loss after an episode of acute rejection.
Levels of mRNA in Urinary Cells
The mean (±standard error [SE]) log-transformed ratio of OX40 mRNA copies to 18S rRNA copies in urinary cells was 6.98±0.26 in the 21 subjects with acute rejection and was substantially higher than that in the 25 subjects with normal biopsy results (stable group) (3.88±0.50, P<0.0001, Mann-Whitney U test, Fig. 1A). The log-transformed value of 18S-normalized OX40L mRNA (3.11±0.40 vs. 0.72±0.41, P=0.0004, Fig. 1B) was also higher in the acute rejection group compared with the stable group. Levels of mRNA for PD-1 (1.29±0.72 vs. −1.47±0.25, P=0.004, Fig. 1C) also differed significantly between the two groups but not the levels of mRNA for PD-L1 (7.87±0.36 vs. 7.03±0.56, P=0.08, Fig. 1D) or PD-L2 (6.27±0.32 vs. 5.49±0.33, P=0.20, Fig. 1E).
ROC Curve Analysis of mRNA Levels for Predicting Acute Rejection
The receiver operating characteristic (ROC) curves (Fig. 2) show the fraction of true-positive results (sensitivity) and false-positive results (1−specificity) for various cutoff levels of an individual mRNA or a combination of mRNAs. The log-transformed cutoff value that gave the maximum combined sensitivity and specificity for OX40 was 5.98; at this cutpoint, the sensitivity was 81%, and the specificity was 88% (AUC 0.88, 95% confidence interval [CI] 0.78–0.98, P<0.0001; Fig. 2A). The cutoff value for OX40L of 1.79 was diagnostic of acute rejection with a sensitivity of 90% and a specificity of 64% (AUC 0.81, 95% CI 0.68–0.94, P<0.0001; Fig. 2B), and the AUC for the combination of OX40 and OX40L was 0.87 (95% CI 0.77–0.98, P<0.0001; Fig. 2C).
By using a stepwise logistic regression analysis, we investigated whether a linear combination of the log-transformed, 18S-normalized levels of urinary cell mRNAs improved the diagnostic accuracy. Our analysis showed that a combination of urinary cell levels of mRNA for OX40, OX40L, PD-1, and Foxp3 increased the AUC to 0.98 (95% CI 0.96–1.00, P<0.0001) and predicted acute rejection with a sensitivity of 95% and a specificity of 92% (Fig. 2D). Inclusion of the remaining four mRNA measures (PD-L1, PD-L2, granzyme B, and perforin) resulted in an AUC of 1.00 (100% sensitivity and 100% specificity), but this model likely reflects overfitting. Importantly, although no significant difference was found between the AUC for the combination of OX40 and OX40L compared with the AUC for OX40 alone (Fig. 2A vs. C; P=0.60), the AUC for the combination of OX40, OX40L, PD-1, and Foxp3 was significantly greater than the AUC for OX40 alone (Fig. 2A vs. D; P=0.03).
Acute Rejection Severity and Urinary Cell mRNA Levels
Within the acute rejection group (n=21), we examined whether acute rejection severity, as reflected by serum creatinine levels, was associated with urinary cell levels of mRNA. Our analysis showed that none of the mRNAs were significantly associated with serum creatinine levels: OX40, Spearman's correlation coefficient (rs)=−0.24, P=0.29; OX40L, rs=−0.05, P=0.83; PD-1, rs=−0.07, P=0.78; PD-L1, rs=0.16, P=0.50; PD-L2, rs=0.02, P=0.95; granzyme B, rs=0.07, P=0.78; perforin, rs=0.05, P=0.84; and Foxp3, rs=−0.04, P=0.85.
Twelve of the 21 acute rejection biopsies were classified as T-cell-mediated acute rejection Banff grade IA and the remaining 9 as Banff grade 1B. Neither serum creatinine level (mean±SE, 2.76±0.48 vs. 3.27±0.46, respectively; P=0.36, Mann-Whitney U test) nor levels of mRNA for OX40 (P=0.46), OX40L (P=0.32), PD-1 (P=0.33), PD-L1 (P=0.27), PD-L2 (P=0.08), granzyme B (P=0.19), perforin (P=0.50), or Foxp3 (P=0.41) were significantly different between the two Banff grades.
Acute Rejection Reversibility and Urinary Cell mRNA Levels
Thirteen of the 21 acute rejections qualified as successfully reversed (16). The mean (±SE) log-transformed level of OX40 mRNA in urinary cells was higher in those with reversible acute rejection than those without reversal (7.47±0.31 vs. 6.18±0.28, P=0.01, Fig. 3A). The levels of OX40L mRNA were also higher in those with reversal than those without reversal (3.88±0.34 vs. 1.87±0.74, P=0.02, Fig. 3B). We also examined whether a relationship exists between log-transformed OX40 and OX40L mRNA levels and percent improvement in 1/creatinine, and this analysis showed that there was no significant linear relationship between the log-transformed levels of OX40 (rs=−0.40, P=0.07) and OX40L mRNA (rs=−0.21, P=0.36) and percent improvement in 1/creatinine.
Acute rejection reversibility was also predicted by the level of mRNA for Foxp3 (P=0.04). In contrast, levels of mRNA for PD-1 (P=0.80), PD-L1 (P=0.33), PD-L2 (P=0.33), granzyme B (P=0.41), and perforin (P=0.21) did not predict acute rejection reversal. In a stepwise logistic regression analysis, Foxp3 was a marginally significant predictor controlling for OX40, and none of the other six mRNA measures was independently predictive of acute rejection reversibility.
ROC Curve Analysis of mRNA Levels for Predicting Acute Rejection Reversal
The log-transformed threshold that had the maximum combined sensitivity and specificity for OX40 mRNA was 6.68, and by using this threshold, reversibility of acute rejection was predicted with a sensitivity of 85% and a specificity of 75% (AUC 0.84, 95% CI 0.66–1.0, P=0.0002) (Fig. 4A). At a cutoff value of 3.79 for OX40L mRNA, reversibility of acute rejection was predicted with a sensitivity of 69% and a specificity of 100% (AUC 0.83, 95% CI 0.65–1.0, P=0.0004; Fig. 4B). The AUC for the combination of OX40 and OX40L was 0.89 (95% CI 0.75–1.0, P<0.0001; Fig. 4C), and the AUC for the combination of OX40 and Foxp3 was 0.90 (95% CI 0.77–1.0, P<0.0001; Fig. 4D). The AUC for the combination of all eight mRNA measures (again an overfitted model) was 0.93 (95% CI 0.80–1.0, P<0.0001). Although the AUCs for the combination of OX40 and OX40L, the combination of OX40 and Foxp3, and the combination of all eight genes were each numerically greater than the AUC of OX40 alone, a statistical comparison of the ROC curves showed that none was significantly greater than the AUC for OX40 alone (all P>0.20).
Although the levels of urinary cell mRNAs for OX40, OX40L, and Foxp3 predicted reversibility of an episode of acute rejection, age (mean [±SD], 41±12.0 years vs. 51±12.7 years; P=0.06), gender (eight men, five women vs. three men, five women; P=0.39), race (two white, four black, and seven other race vs. two white, three black, and three other race; P=0.75), donor type (nine live and four deceased vs. four live and four deceased; P=0.65), or the type of acute rejection therapy (nine with single modality of treatment and four with combined modalities vs. seven with single modality of treatment and one with combined modalities; P=0.61) were not predictive. Also, the creatinine level at the time of biopsy was not predictive of acute rejection reversal (2.79±1.19 vs. 3.28±2.05, P=0.64).
Graft Loss and Urinary Cell mRNA Levels
Of the 21 subjects with acute rejection, 5 lost their grafts within 6 months of biopsy-proven acute rejection. The mean creatinine [±SE] at the time of biopsy of the five patients with graft loss was 4.14±0.98 mg/dL and higher than those without graft loss (2.61±0.29 mg/dL), but this difference was not significant (P=0.12).
In the logistic regression analysis, OX40 was the mRNA measure that had the strongest association, albeit not statistically significant (P=0.17), with graft loss; those with lower OX40 were more likely to lose their graft. The log-transformed threshold that had the maximum combined sensitivity and specificity for OX40 mRNA was 6.68, and by using this threshold, graft loss was predicted with a sensitivity of 80% and a specificity of 75% (AUC 0.74, 95% CI 0.51–0.97, P=0.04; Fig. 5A), and the AUC for OX40L was 0.68 (95% CI 0.44–0.92, P=0.14; Fig. 5B). The AUC for the combination of OX40 and OX40L was 0.75 (95% CI 0.52–0.98, P=0.04; Fig. 5C). However, there was no significant difference between the AUC for OX40 alone and the AUC for the combination of OX40 and OX40L (Fig. 5A vs. C; P=0.48). The difference between the AUC for OX40 alone and the AUC for the combination of all eight genes was marginally significant (P=0.09), but this prediction model is confounded by the inclusion of a relatively large number of mRNA measures for the small number of outcome events observed.
We have identified that acute rejection of human renal allografts, an important posttransplant complication, is associated with the levels of mRNA for OX40, OX40L, and PD-1 in urinary cells and that a diagnostic algorithm that included urinary cell levels of mRNA for OX40, OX40L, PD-1, and Foxp3 is an excellent predictor of acute rejection of human renal allografts. Our urinary cell mRNA profiling strategy also indicates that acute rejection reversal and perhaps graft loss after an episode of acute rejection are predictable with the use of levels of mRNAs in urinary cells, but these results need to be confirmed in a larger study. It is noteworthy that OX40 was the strongest predictor of all three outcomes and that Foxp3 was the second strongest predictor of both acute rejection and acute rejection reversal.
OX40 is a member of the tumor necrosis factor receptor superfamily, expressed primarily by CD4+ T cells and some CD8+ T cells. The ligand for OX40 is typically expressed by antigen-presenting cells and is also a member of the tumor necrosis factor superfamily. Stimulation through the OX40/OX40L pathway propagates T-cell expansion, cytokine production, and generation of memory T cells. OX40 costimulation typically evokes a T helper 2 response signal; however, it is also capable of a T helper 1 response (17–21). Elegant preclinical studies showed that early OX40 stimulation results in acute rejection, whereas delayed stimulation leads to chronic rejection of experimental allografts (9). The new observation from this study that the levels of transcripts for OX40 and OX40L are higher during an episode of acute rejection compared with stable allograft status is consistent with the preclinical findings that the OX40/OX40L pathway facilitates acute rejection. Importantly, our novel observations demonstrate for the first time that noninvasively ascertained levels of mRNA for OX40 and OX40L are positively associated with an episode of acute rejection in humans.
The OX40/OX40L proteins have also been implicated in regulating the emergence of Tregs and modulating their suppressive activity (17, 21). Our observation that acute rejection reversal is associated with high rather than low levels of OX40 seems to conflict with the notion that OX40 signaling may impair Treg-associated down-regulation of the antiallograft response. However, it seems that the role of OX40 in Treg-associated processes is complex and not unidirectional. Griseri et al. (22) recently reported that the presence of OX40 on the surface of Tregs is an obligatory requirement for the Tregs to accumulate at the site of inflammation and control colitogenic effector T cells. In their experimental model of colitis, OX40 was also required for the effector response. This dual role for OX40-positive cells, one protective and the other pathogenetic, may account for our finding that high OX40 is associated not only with acute rejection but also with acute rejection reversal. Our new finding is also reminiscent of our earlier observation that Foxp3 is overexpressed during an episode of acute rejection and that acute rejection reversal is associated with high rather than low levels of Foxp3 (16).
A novel T-cell inhibitory pathway involves PD-1 and its ligands, PD-L1 and PD-L2 (23–25). The PD-1 and PD-L interaction is reported to arrest the cell cycle in the G0 to G1 phase (25), but PD-1 activation resulting in T-cell proliferation has also been reported (26). It has also been reported that blockade of PD-1 or PD-L1 results in accelerated acute rejection in preclinical models (14, 25, 27). Our findings suggest that the urinary cell levels of PD-1, PD-L1, and PD-L2 are less informative of allograft status compared with the levels of mRNA for the positive costimulatory proteins OX40 and OX40L.
Measurement of OX40 mRNA and OX40L mRNA in urinary cells offers a noninvasive means of diagnosing acute rejection and accurately predicting the outcome of an episode of acute rejection. The current investigation confirmed and extended our earlier finding that urinary cell levels of Foxp3 predict acute rejection reversal (16). We also found that a combination of levels of mRNA for OX40 and Foxp3 resulted in a larger AUC compared with the AUC of OX40, or the combination of OX40 and OX40L but the null hypothesis that the improvement in AUC was due to chance alone could not be rejected in this study.
In this study, urinary cell level of OX40 mRNA was the strongest predictor of graft loss, and none of the other mRNAs were significant predictors. Inclusion of all eight mRNAs in a prediction model resulted in the largest AUC that was marginally better than the AUC of OX40 alone. This result, however, should be interpreted with due caution in view of the small sample size and the potential for model overfitting.
In summary, our findings demonstrate that measurement of mRNA for OX40, OX40L, and PD-1 in urine offers a noninvasive means of diagnosing acute rejection of human renal allografts, and a combination of urinary cell levels of mRNA for OX40, OX40L, PD-1, and Foxp3 is the best predictor of biopsy-confirmed acute rejection. Among the mRNAs measured, OX40 mRNA seemed to outperform others with respect to predicting acute rejection reversibility and graft loss after an episode of acute rejection, and studies to validate the biomarkers of renal allograft status discovered in this study merit further consideration.
MATERIALS AND METHODS
We examined urine samples from 46 kidney transplant recipients. There were 21 subjects with graft dysfunction (mean±SE creatinine level 2.98±0.34 mg/dL) and biopsy-confirmed acute rejection (mean±SD age, 45±13 years; 11 men and 10 women; 4 white, 7 black, and 10 other race; 8 deceased donor and 13 living donor grafts; median time from transplantation to for-cause biopsy, 180 days) and 25 subjects with stable graft function (creatinine level 1.40±0.06 mg/dL) and normal biopsy results (age 47±13 years; 14 men and 11 women; 9 white, 5 black, and 11 other race; 5 deceased donor and 20 living donor grafts; time from transplantation to protocol biopsy, 51 days). Forty-four of the 46 urine specimens were collected on the day of the biopsy. The remaining two samples were from the normal biopsy group and collected before the biopsy procedure.
Formalin-fixed, paraffin-embedded renal biopsy specimens were stained with hematoxylin-eosin, periodic acid-Schiff, and Masson's trichrome stains and were scored on the Banff classification (28) by a pathologist blinded to the results of the gene expression. There was no evidence of BK virus or cytomegalovirus on light microscopy. Additional information regarding the study subjects is provided as Supplemental Digital Content (see Supplemental Digital Content 1, https://links.lww.com/TP/A293).
Quantitation of mRNA by Real-Time Quantitative Polymerase Chain Reaction Assays
Total RNA was isolated from urine cell pellets, quantified, and reverse transcribed to cDNA. Polymerase chain reaction analysis, quantification of mRNA copies, and the sequence of the gene-specific oligonucleotide primers and TaqMan probes used in this investigation are described in Supplemental Table 1 (see Supplemental Digital Content 2, https://links.lww.com/TP/A294).
The levels of 18S-normalized mRNA for OX40, OX40L, PD-1, PD-L1, PD-L2, granzyme B, perforin, and Foxp3 deviated from a normal distribution (P<0.0001) and natural log-transformation substantially reduced the deviation. All statistical calculations were performed using GraphPad Prism software version 4.0 (GraphPad Software, Inc. La Jolla, CA). By using 18S-normalized levels, we used the Mann-Whitney U test to test the differences between the group with acute rejection and the group with normal biopsy results. Categorical variables were compared using Fisher's exact test or chi-square analysis. We used Spearman's rank-order correlations to test for monotonic associations with the 18S-adjusted mRNA transcript levels, and Supplemental Table 2 (see Supplemental Digital Content 3, https://links.lww.com/TP/A295) shows the relationships among the levels of mRNAs measured in this investigation. We generated ROC curves for individual mRNA levels and a linear combination of mRNA levels to determine the cutoff points that yielded the highest combined sensitivity and specificity for predicting the diagnosis, reversibility of an episode of acute rejection, and graft loss after acute rejection. Successful reversal of acute rejection was defined as a return in creatinine to within 15% of baseline within 4 weeks of initiating antirejection treatment (16). A second endpoint was the loss of the graft during the first 6 months after the diagnosis of acute rejection.
1. Hariharan S, Johnson CP, Bresnahan BA, et al. Improved graft survival after renal transplantation in the United States, 1988 to 1996. N Engl J Med
2000; 342: 605.
2. Bustami RT, Ojo AO, Wolfe RA, et al. Immunosuppression and the risk of post-transplant malignancy among cadaveric first kidney transplant recipients. Am J Transplant
2004; 4: 87.
3. Jamil B, Nicholls K, Becker GJ, et al. Impact of acute rejection therapy on infections and malignancies in renal transplant recipients. Transplantation
1999; 68: 1597.
4. Cornell LD, Smith RN, Colvin RB. Kidney transplantation: Mechanisms of rejection and acceptance. Annu Rev Pathol
2008; 3: 189.
5. Rothstein DM, Sayegh MH. T-cell costimulatory pathways in allograft rejection and tolerance. Immunol Rev
2003; 196: 85.
6. Simeoni L, Lindquist JA, Smida M, et al. Control of lymphocyte development and activation by negative regulatory transmembrane adapter proteins. Immunol Rev
2008; 224: 215.
7. Bharat A, Mohanakumar T. Allopeptides and the alloimmune response. Cell Immunol
2007; 248: 31.
8. Game DS, Lechler RI. Pathways of allorecognition: Implications for transplantation tolerance. Transpl Immunol
2002; 10: 101.
9. Burrell BE, Lu G, Li XC, et al. OX40 costimulation prevents allograft acceptance induced by CD40-CD40L blockade. J Immunol
2009; 182: 379.
10. Chen M, Xiao X, Demirci G, et al. OX40 controls islet allograft tolerance in CD154 deficient mice by regulating FOXP3+
2008; 85: 1659.
11. Demirci G, Amanullah F, Kewalaramani R, et al. Critical role of OX40 in CD28 and CD154-independent rejection. J Immunol
2004; 172: 1691.
12. Fife BT, Pauken KE, Eagar TN, et al. Interactions between PD-1 and PD-L1 promote tolerance by blocking the TCR-induced stop signal. Nat Immunol
2009; 10: 1185.
13. Gao W, Demirci G, Strom TB, et al. Stimulating PD-1-negative signals concurrent with blocking CD154 co-stimulation induces long-term islet allograft survival. Transplantation
2003; 76: 994.
14. Wang W, Carper K, Malone F, et al. PD-L1/PD-1 signal deficiency promotes allogeneic immune responses and accelerates heart allograft rejection. Transplantation
2008; 86: 836.
15. Li B, Hartono C, Ding R, et al. Noninvasive diagnosis of renal-allograft rejection by measurement of messenger RNA for perforin and granzyme B in urine. N Engl J Med
2001; 344: 947.
16. Muthukumar T, Dadhania D, Ding R, et al. Messenger RNA for FOXP3 in the urine of renal-allograft recipients. N Engl J Med
2005; 353: 2342.
17. Demirci G, Li XC. Novel roles of OX40 in the allograft response. Curr Opin Organ Transplant
2008; 13: 26.
18. Gramaglia I, Jember A, Pippig SD, et al. The OX40 costimulatory receptor determines the development of CD4 memory by regulating primary clonal expansion. J Immunol
2000; 165: 3043.
19. Li XC, Rothstein DM, Sayegh MH. Costimulatory pathways in transplantation: Challenges and new developments. Immunol Rev
2009; 229: 271.
20. Takeda I, Ine S, Killeen N, et al. Distinct roles for the OX40-OX40 ligand interaction in regulatory and nonregulatory T cells. J Immunol
2004; 172: 3580.
21. Xiao X, Kroemer A, Gao W, et al. OX40/OX40L costimulation affects induction of Foxp3+
regulatory T cells in part by expanding memory T cells in vivo. J Immunol
2008; 181: 3193.
22. Griseri T, Asquith M, Thompson C, et al. OX40 is required for regulatory T cell-mediated control of colitis. J Exp Med
2010; 207: 699.
23. del Rio ML, Buhler L, Gibbons C, et al. PD-1/PD-L1, PD-1/PD-L2, and other co-inhibitory signaling pathways in transplantation. Transpl Int
2008; 21: 1015.
24. Kitazawa Y, Fujino M, Wang Q, et al. Involvement of the programmed death-1/programmed death-1 ligand pathway in CD4+
regulatory T-cell activity to suppress alloimmune responses. Transplantation
2007; 83: 774.
25. Sandner SE, Clarkson MR, Salama AD, et al. Role of the programmed death-1 pathway in regulation of alloimmune responses in vivo. J Immunol
2005; 174: 3408.
26. Tamura H, Dong H, Zhu G, et al. B7-H1 costimulation preferentially enhances CD28-independent T-helper cell function. Blood
2001; 97: 1809.
27. Ito T, Ueno T, Clarkson MR, et al. Analysis of the role of negative T cell costimulatory pathways in CD4 and CD8 T cell-mediated alloimmune responses in vivo. J Immunol
2005; 174: 6648.
28. Racusen LC, Solez K, Colvin RB, et al. The Banff 97 working classification of renal allograft pathology. Kidney Int
1999; 55: 713.