Intragraft Levels of Foxp3 mRNA Predict Progression in Renal Transplants with Borderline Change

Abstract

The optimal therapeutic management of borderline lymphocytic infiltrates in renal allografts, described by Banff criteria, is unknown, largely because of the inability to predict clinical outcome in these cases. For determination of molecular factors that may predict outcome in cases of borderline change histology, mRNA levels of Foxp3, Granzyme B, IFN-γ, IL-23, and RORγt were measured in renal tissue from 46 untreated patients. Twenty-five patients were considered “nonprogressive,” defined by a serum creatinine that remained <110% of baseline during the 40 d after biopsy. Twenty-one patients were considered “progressive,” defined by an increase in serum creatinine >110% from baseline and by repeat histologic examination within 40 d showing progression toward acute rejection. Only Foxp3 mRNA levels were significantly higher in nonprogressors than in progressors (P = 0.001). Analysis of receiver operating characteristic curves demonstrated that the outcome for patients with biopsies showing borderline change could be predicted with 90% sensitivity and 79.1% specificity using the optimal Foxp3 mRNA cutoff value. Our findings suggest that the measurement of Foxp3 mRNA offers a means of improving prediction of outcome of borderline change.

The diagnosis and treatment of acute allograft rejection are crucial determinants of long-term graft outcome and therefore present important challenges in the field of kidney transplantation. The interpretation of histologic analyses relies on the widely used Banff working classification of renal pathology. The Banff scheme defines the minimum threshold for T cell–mediated acute rejection (TCMR) and refers to borderline change (BL) biopsies as changes insufficient for diagnosis of acute rejection, including foci of tubulitis (t1, t2, or t3) with mild to moderate (<25%) cortical infiltration (i0 or i1) and without intimal arteritis.¹ Although the Banff classification is helpful for clinical treatment of transplant recipient with TCMR, there are currently no guidelines for the therapeutic treatment of patients with BL changes. Some studies conclude that most BL cases are indeed TCMR,² but discrepant percentages (30 to 80%) of BL cases have been reported to progress to TCMR without treatment.^3,4 This may be explained by the “single-cell cutoff” concept applied in histologic classification⁵ and highlights the need for new technology such as transcriptomics to be used to analyze the immunologic features of T cell infiltrates and to predict clinical outcome. We previously demonstrated that the activity levels of the cytotoxic molecules Granzyme B (GB) and Fas ligand and the regulatory T cell (Treg) marker Foxp3, from the intragraft infiltrate of biopsies with BL changes, are intermediate, between those associated with TCMR and absence of rejection.^6,7 We also quantified the Foxp3/GB ratio to determine the relative proportions of alloaggressive and graft-protecting T cells. T cell infiltrate from biopsies with BL changes had a significantly higher ratio than in grade IA acute rejection biopsies, although the distribution of both markers was heterogeneous in all groups.⁷

In 2005, Muthukumar et al.⁸ showed that Foxp3 mRNA levels in the urine of patients with acute rejection (AR) predicts the reversal of AR and identifies patients at risk from graft failure within 6 mo. In this study, we analyzed the ability of intragraft mRNA levels for Foxp3, GB, INF-γ, and two molecular markers of the Th17 pathway, RORγt and IL-23, to predict the outcome (progressive or nonprogressive) of patients with untreated BL changes.

We first analyzed Foxp3, GB, ΙΝFγ, RORγt and IL-23 mRNA levels in 46 samples from patients with untreated BL changes (Figure 1A). The expression level mean ratio of Foxp3, GB, INF-γ, IL-23, and RORγt to glyceraldehyde-3-phosphate dehydrogenase (GAPDH) were respectively 0.048 (±0.086), 0.629 (±1.790), 0.489 (±1.450), 5.160 (±11.520), and 0.743 (±0.728). Intragraft expression of Foxp3 in renal biopsy from a patient with BL change in Cd4⁺ T cells is depicted in Figure 1B.

Figure 1.

(A) Expression level of five genes (encoding Foxp3, GB, IFN-γ, IL-23, and RORγt) assessed in biopsy samples from 46 BL patients. (B) Detection of FoxP3⁺ T cells in renal biopsies using double CD4⁺ (green) Foxp3 (red) immunostaining in a patient with BL change from the NP group. Magnification, ×100.

Second, we analyzed mRNA levels for each of these factors as a function of clinical outcome of patients with BL changes. Patients were assigned to either a nonprogressive (NP) or a progressive (P) phenotype according to clinical criteria. The NP group included patients who did not progress to an “acute rejection” phenotype, defined in all cases by serum creatinine value <110% of the baseline at 40 d after detection of BL changes; however, because BL change was diagnosed in nine of these patients with suboptimal renal function (unsatisfactory improvement in serum creatinine), this clinical criterion was judged as insufficient and completed by a second renal biopsy performed within 40 d after detection of BL changes showing an absence of progression toward acute rejection. The P group was also defined by clinical criteria (serum creatinine >110% baseline), but P phenotype was confirmed in all cases by a second renal allograft biopsy performed within 40 d showing progression toward acute rejection defined by tubulointerstitial inflammatory indices of i2 and t2 or greater or vascular index v1 or greater.

According to these criteria, the outcome of 25 (54%) untreated patients was considered NP, defined by a reduction in serum creatinine to <110%. Because diagnosis of BL changes was performed in nine of the cases because of suboptimal renal function (unsatisfactory improvement in serum creatinine) a second renal biopsy was performed showing the absence of acute rejection in all cases (seven cases with biopsy showing persistent BL change and two without significant T cell infiltrate). Twenty-one (46%) patients were considered P on the basis of clinical criteria (serum creatinine >110% baseline) and confirmed by histologic examination performed on all patients within 40 d. Histologic examination included grade IA acute rejection in 11 patients, grade IB in seven patients, and grade IIA in three patients, according to Banff classification. These patients all received antirejection therapy based on intravenous administration of high-dosage glucocorticoids.

GB, Foxp3, INF-γ, IL-23, and RORγt mRNA levels were assessed in both NP and P groups (Figure 2). Only Foxp3 mRNA levels were significantly higher (P = 0.001) in the NP group than in the P group. The median values in the NP and P groups, respectively, were as follows: Foxp3 0.048 and 0.000 (P = 0.001); GB 0.115 and 0.840 (P = 0.85); IFN-γ 0.221 and 0.010 (P = 0.06); IL-23 1.420 and 0.687 (P = 0.09); and RORγt 0.525 and 0.372 (P = 0.06).

Figure 2.

(A through E) mRNA levels for Foxp3 (A), GB (B), IFN-γ (C), IL-23 (D), and RORγt (E) in P and NP groups.

We analyzed the ability of a Foxp3 mRNA diagnostic test to distinguish between the NP and P groups (Figure 3). Receiver operating characteristic (ROC) curve analysis shows the proportion of true-positive results (sensitivity) and false-positive results (1 − specificity) for Foxp3 mRNA levels. The calculated area under the curve was 0.86 (95% confidence interval 0.75 to 0.9). Using the cutoff value of >0.0115 derived from the data, the Foxp3 mRNA level was predictive of a good prognosis, with a sensitivity of 79% and a specificity of 90%. These results clearly suggest that Foxp3 is a powerful marker to distinguish between the two groups. At 1 yr, mean serum clearance was 42.9 (±16.3) and 40.5 (±13.5) in the NP and P groups, respectively (P = 0.6). At 2 yr, mean serum clearance was 43.7 (±19.5) and 41.6 (±15.5) in the NP and P groups, respectively (P = 0.7). Although time from transplantation to BL change was higher in the P group, the difference did not reach statistical significance (P = 0.09).

Figure 3.

ROC curve for Foxp3 mRNA levels as predictor of NP outcome.

A previous study demonstrated that Foxp3 levels may serve as a mechanistically informative biomarker of acute rejection.⁸ More recently, significant levels of Foxp3 were detected in biopsies from tolerant recipients after combined bone marrow and kidney transplants.⁹ Here, we demonstrate for the first time that intragraft Foxp3 mRNA levels and Foxp3/GB ratio can distinguish between a P and an NP outcome after diagnosis of BL changes. These findings are consistent with the hypothesis that Treg play a role in “damage control” during T cell infiltration after allograft transplantation. We suggest that, as in TCMR, BL changes involve the activation of both cytotoxic and Treg and that the transition of this infiltrate into a more aggressive phenotype or its spontaneous regression is, in part, related to the number of Treg and ratio of Treg/cytotoxic T cells. In contrast, neither GB nor INF-γ seemed to be related to outcome, highlighting the master role of Treg in human transplantation.

IL-23, a member of the IL-12 cytokine family, stimulates the production of IL-17 in a CD4⁺ T cell population named Th17 and RORγt, an orphan nuclear receptor, which promotes mouse and human Th17 differentiation.¹⁰ IL-23–mediated activation of Th17 is a major event in autoimmunity,^11,12 but its role in transplantation immunity is unknown. Th17 cells also share a complex relationship with Treg and thus may affect Treg-induced tolerance: Naive CD4⁺ T cells differentiate into Treg when exposed to TGF-β¹³ but differentiate into Th17 when exposed to both TGF-β and IL-6.¹⁴ We analyzed IL-23 and RORγt mRNA levels in patients with BL changes to determine the relative abundance of Th17/Treg populations but did not find a significant correlation between these mRNA levels and clinical outcome.

The detection of infiltrate with BL changes seems to have a number of implications for renal transplantation. It has been found in nearly 20% of cases with allograft dysfunction^3,4 and in up to 35% of biopsy specimens from transplant patients with stable renal function.^15,16 Paradoxically, no therapeutic guidelines for treating BL change infiltrates are available. Several uncontrolled studies have retrospectively addressed the effects of antirejection therapy, generating controversial results.^2,3,17 Moreover, clinical outcomes of the patients have been little studied.¹⁸ In this study, we included only patients with untreated BL change.

Because recommendations to treat or not to treat BL change are not currently clarified, therapeutic decisions depend in most cases on subjective and unvalidated criteria, including (1) time to diagnosis of the borderline infiltrate (late diagnoses were considered indicative of a bad prognosis and were frequently treated) and (2) histologic infiltrate score (patients with t2 and t3 infiltrate were more likely to be treated, whereas those with t1 T cell infiltrate were likely to be untreated). Associated immunologic risk factors have also influenced therapeutic decisions, because patients were more likely to be treated when they had received a second transplantation or had a significant level of anti-HLA panel reactive antibodies. Thus, decision to treat or not BL changes is likely to be related to empirical criteria, which could introduce a selection bias in our group.

Our results and previous findings⁴ suggested that BL change infiltrate is a heterogeneous entity. We suggest that the level of intragraft Foxp3 mRNA is an informative biomarker for BL outcome. This could lead to future individualized treatment of such patients, for example, in determining the decision to give antirejection therapy or to perform subsequent biopsy in patients with low Foxp3 mRNA levels. Similar analyses need to be performed in BL occurring in patients with stable renal function (subclinical BL changes). It was recently demonstrated that Foxp3 could be transiently induced by activating CD4⁺CD25⁻ T cells through T cell receptor cross-linking, but suppressor activity from these cells was very low or absent.¹⁹ This contrasts with the suppressor activity of the CD4⁺ activated T cells with constitutive Foxp3 levels. Overall, these data suggest that Foxp3 is not produced exclusively in Treg and that analysis of Foxp3 in clinical situations such as alloreactivity should be interpreted with caution. Consequences of Foxp3 mRNA levels on long-term graft function also need to be determined.

CONCISE METHODS

Patients

Between 2002 and 2006, 346 transplants were performed in our transplant department. Diagnosis of BL change was established in 28% (n = 97) of renal biopsies performed for allograft dysfunction or unsatisfactory improvement in serum creatinine. Among this allograft kidney biopsy collection, 46 were considered eligible for our study on the following criteria: Patients with untreated BL changes (histologic indices i1t1 and i1t2) occurring in the early phase after transplantation (<180 d), with frozen renal-core biopsy tissue available, were selected. Immunostaining for C4d was negative in all specimens. At the time of detection of BL changes, the mean (SD) increase in serum creatinine from baseline (mean of the lowest two consecutive values) was 35 (±26) μmol/L. The median interval between renal transplantation and biopsy was 44.5 d. Kidney allograft biopsies were obtained from all graft recipients to determine the cause of graft dysfunction, including unsatisfactory improvement in serum creatinine or rise in serum creatinine >10% above baseline. Clinical data are depicted in Table 1 according to the P or NP phenotype.

Table 1.

Clinical data according to the P and NP phenotypes^a

Parameter	NP Group(n = 25)	P Group(n = 21)	P
Age of recipient (yr; mean ± SD)	46.2 ± 1.88	44.00 ± 1.60	0.30 (NS)
Cold ischemia time (h; mean ± SD)	25.4 ± 1.40	26.19 ± 1.60	0.58 (NS)
HLA antibodies before transplantation (%)	24	17	0.25 (NS)
No. of HLA mismatch (mean ± SD)	2.25 ± 0.14	2.50 ± 0.16	0.70 (NS)
Induction therapy			0.70 (NS)
IL-2R	19	16
ATG	5	3
no induction	1	2
Maintenance of IS regimen			0.20 (NS)
CsA + AZA or MMF + steroids	5	6
Tac + AZA or MMF + steroids	20	15
Delayed graft function (%)	35	32	0.90 (NS)
Time to BL changes (d; median)	36.52	52.20	0.09 (NS)

^aATG, antithymocyte globulin; MMF, mycophenolate mofetil; CsA, cyclosporine; Tac, tacrolimus.

RNA Isolation and Quantification of Gene Expression by Real-Time Quantitative Reverse Transcriptase–PCR

We investigated intrarenal mRNA levels for GB, Foxp3, INF-γ, RORγt, and IL-23 in the 46 selected biopsies. Total RNA tissue was purified using the RNeasy kit (Qiagen SA, Courtaboeuf, France). For RNA quantity, we used nanotrope technology. For assessment of the integrity of total RNA extracted, RNA was checked using the RNA 6000 Pico Assay kit combined with the Agilent 2100 Bioanalyser and the RNA Integrity Number was calculated by Agilent Bioanalyser Expert software. All cDNA analyzed were measured by relative quantification using real-time PCR. Quantitative PCR was performed in an ABI prism 7900 using a SYBR Green. The GAPDH housekeeping gene was chosen as a control, and its expression levels were used for normalization of data. GAPDH mRNA levels were similar for all groups (data not shown). Primer and probe sequences and PCR parameters are indicated in Table 2. First-strand cDNA was synthesized in reverse transcriptase samples, containing total RNA isolated from biopsies, 16 U/μl M-MLV reverse transcriptase (Life Technologies-BRL, Life Technologies, Cergy-Pontoise, France), 4 μM Oligo-(dT) 12-18 (Amersham-Pharmacia Biotech, Saclay, France), and 0.8 mM mixed dNTP (Amersham-Pharmacia Biotech, Saclay, France). cDNA dilution series of calibrator was used to set up a standard curve for target genes and the housekeeping gene; Ct values were plotted against the log cDNA concentration added. We used the resulting linear graphs to determine the differences in Ct values for each sample, expressed as a relative percentage of mRNA present in the reference calibrator dilution, according to the ΔΔCt method (2^−ΔΔCt). Only samples meeting the following criteria were considered for analysis: (1) Housekeeping gene expressed at a sufficient level (crossing point <31 cycles), and (2) at least one other cDNA analyzed (GB, Foxp3, INF-γ, RORγt, and IL-23) was detected.

Table 2.

Sequences of oligonucleotide primers and FRET probes used for the quantification of target genes by quantitative PCR

Gene	Sequence Reference	Sense	Sequence	Amplicon
FoxP3	NM_014009	5′	TCC ACA ACA TGC GAC CCC CTT TCA	217
		3′	ACA GCC CCC TTC TCG CTC TCC A
GB	NM_004131	5′	GGG GAA GCT CCA TAA ATG TCA CCT TG	216
		3′	GCT TCA CCT GGG CCT TGT TGC TAG
RORγ t	NM_005060 and NM_001001523.1	5′	CTC CAT CTT TGA CTT CTC CCA CTC CCT A	247
		3′	CAC ATG CTG GCT ACA CAG GCT C
IFN-γ	NM_000619	5′	TGG CTT AAT TCT CTC GGA AAC G	177
		3′	AAA AGA GTT CCA TTA TCC GCT ACA TC
IL-17	NM_002190	5′	AGG CAG GAA TCA CAA TCC CAC GA	244
		3′	TCA GCG TTG ATG CAG CCC AAG T
IL-23	NM_016584	5′	TTC TCT GCT CCC TGA TAG CCC TGT G	147
		3′	CGG AGA AGG AGA CGC TGC CA

Detection of Foxp3 in CD3⁺ T Cells from Renal Biopsy

Frozen tissue samples were stained with polyclonal goat anti-human FoxP3 (ab2481, 500 μg/ml, IgG, 1:100 dilution; Abcam, Cambridge, UK) and mouse anti-human CD4 (MT310, IgG1, 100 μg/ml, 1:100 dilution; DakoCytomation, Carpinteria, CA), followed by FITC-conjugated rat anti-mouse (145-095-166, IgG, 1 mg/ml, 1:100 dilution; Jackson Immunoresearch, West Grove, PA) and biotinylated rabbit anti-goat (E0466, 1 mg/ml, 1:400 dilution; DakoCytomation) followed by cyanine 3–conjugated streptavidin (PA43001, 1 mg/ml, 1:300 dilution; Amersham Biosciences, Uppsala, Sweden).

Statistical Analysis

Results are presented as median and interquartile range. Nonparametric tests were used for non-normal distributions of mRNA levels. Differences in immune markers between the two groups (P and NP groups) were analyzed by the Mann-Whitney test. The NP and P groups were defined by clinical criteria described already. The P phenotype was systematically confirmed by a second histologic examination performed within 40 d after episode of BL change. We used ROC curves and calculated the area under the curve to analyze the potential diagnostic value of Foxp3 mRNA levels for the prediction of P or NP outcome. The sensitivity and specificity for various cutoff values are given, with their 95% confidence intervals. The cutoff of Foxp3 for the ROC analysis was calculated using the Youden index to choose the optimal cutoff associated with the best specificity. Statistical analyses were performed using SPSS 12.0 (SPSS, Chicago, IL) and BMDP statistical software (Statistical Solutions, Ltd., Cork, Ireland).

DISCLOSURES

None.