Long-Term Kinetics of Intragraft Gene Signatures in Renal Allograft Tolerance Induced by Transient Mixed Chimerism

Abstract

Background.

Renal allograft tolerance has been successfully induced in nonhuman primates (NHPs) and humans through the induction of transient mixed chimerism. To elucidate the mechanisms of tolerance, we compared local immunological responses in renal allografts with those in T-cell mediated rejection (TCMR) and chronic antibody-mediated rejection (CAMR) in NHPs.

Methods.

Using the NanoString nCounter platform, we retrospectively studied 52 mRNAs in 256 kidney allograft samples taken from NHP kidney recipients of donor bone marrow transplantation (DBMT). No immunosuppression was given after 1-month post-DBMT. Recipients that achieved tolerance (TOL, n=13) survived for >1840 ± 1724 days with normal kidney function, while recipients with CAMR (n=13) survived for 899 ± 550 days with compromised graft function, and recipients with TCMR (n=15) achieved only short-term survival (132 ± 69 days).

Results.

The most prominent difference between the groups was FOXP3, which was significantly higher in TOL than in CAMR and TCMR, both early (<1 year p<0.01) and late (≥1 year p<0.05) after transplant. Other mRNAs related to Tregs, such as IL10, TGFB, and GATA3, were also high in TOL. In contrast, transcripts of inflammatory cytokines were higher in TCMR, while activated endothelium associated transcripts were higher in CAMR than in TOL. The ROC analyses revealed that intragraft FOXP3 and CAV1 can reliably distinguish TOL from CAMR.

Conclusions.

High FOXP3 and other Treg-related-mRNAs together with suppressed inflammatory responses and endothelial activation in renal allografts, suggest that intragraft enrichment of Tregs is a critical mechanism of renal allograft tolerance induced by transient mixed chimerism.

INTRODUCTION

Induction of allograft tolerance remains the ultimate goal in organ transplantation, both to eliminate the complications associated with life-long immunosuppression and to prevent chronic rejection.^1,2

Drawing on a series of murine studies performed by Sharabi,^3,4 we previously developed a conditioning regimen that promotes renal allograft tolerance in non-human primates (NHPs) through the induction of transient mixed chimerism.^5–11 This approach has been successfully translated to human leukocyte antigen (HLA)-mismatched kidney transplant (KTx) recipients, who, to date, have achieved immunosuppression-free allograft survival, currently exceeding 15 years.^1,12,13,14 In contrast to skin allograft tolerance achieved in murine experiments, in which permanent mixed chimerism is required for stable tolerance, renal allograft tolerance achieved in NHPs and humans was induced by only transient mixed chimerism. Thus ongoing thymic deletion of donor reactive T cells would not appear to be the mechanistic pathway to tolerance in the primate recipients.³ Instead, our recent in vitro studies in NHPs reveal that TGFβ-dependent, peripheral donor-specific Tregs (pTregs) converted from non-Tregs play an important role in renal allograft tolerance in our model.¹⁵ The importance of the renal allograft itself in the induction/maintenance of tolerance has also been demonstrated in our combined heart and kidney transplant model in NHPs, where, in addition to transient mixed chimerism, co-transplantation of the renal allograft is required to induce heart allograft tolerance.¹⁶ Our clinical trials of tolerance induction detected by quantitative reverse transcription polymerase chain reaction (qRT-PCR) significantly higher FOXP3 with low GZMB mRNA levels in renal allograft biopsies from recipients who achieved tolerance, compared with renal allograft biopsies taken from stable recipients who had remained on conventional immunosuppression.¹² However, the kinetics and stability of these intra-renal genes could not be evaluated because of the limited number of biopsies performed in these human recipients. This emphasizes that a long-term study that observes each recipient longitudinally is necessary to better understand the local immune responses in the renal allografts.

A novel high throughput gene expression analysis platform, NanoString nCounter, has recently been developed that utilizes a barcode-labeled probe-based methodology to detect the expression of up to 800 genes in a single hybridization.¹⁷ A major advantage of this platform includes the ability to work reliably with archival formalin-fixed and paraffin-embedded (FFPE) tissue samples.^18–20 This enabled us to retrospectively analyze various gene signatures in our archive of FFPE renal allograft samples, representing all NHP recipients of combined kidney and bone marrow transplantation (CKBMT) at our institution over the last 25 years. Our initial study identified three dominant factors relating to tolerance or acute/chronic rejection.²¹ In the current study, to further elucidate local immune responses in the renal allografts, we analyzed long-term kinetics of 52 genes, which are relevant to alloimmune responses, rejection and tolerance^{12,19,20,22–31}, in 42 recipients categorized as tolerant (TOL), chronic antibody mediated rejection (CAMR), or T cell mediated rejection (TCMR).

MATERIALS AND METHODS

Animals

Cynomolgus monkeys weighing 4 - 7 kg were used (Charles River Primates, Wilmington, MA, USA). All surgical procedures and postoperative care of animals were performed in accordance with National Institute of Health guidelines for the care and use of primates and were approved by the Massachusetts General Hospital Institutional Animal Care and Use Committee.

Cynomolgus MHC genotyping

Recipient and donor pairs were selected for compatible ABO blood types and mismatched cynomolgus leukocyte major histocompatibility complex (MHC) antigens. MHC characterization was performed as previously described.^32,33 Briefly, genomic DNA was prepared from PBMCs and splenocytes. Panels of 17 microsatellite loci spanning ~5 Mb of the MHC region were amplified from the genomic DNA with fluorescent-labeled PCR primers, and fragment size analysis was determined. The microsatellite haplotypes for each animal were converted to predicted MHC genotypes based on previous cloning and sequencing work with cynomolgus monkeys. The MHC class I (A and B loci) and II (DP, DQ and DR loci) gene disparities in each recipient and donor pairs are shown in Supplemental Table 1.

Kidney and donor bone marrow transplantation protocol

The recipients received donor bone marrow transplantation (DBMT) either simultaneously or several months after KTx after treatment with a non-myeloablative conditioning regimen (Figure 1A). The conditioning protocols for DBMT include low dose total body irradiation (TBI, 1.5 Gy × 2 on days −6 and −5), thymic irradiation (TI, 7 Gy on day −1), pre-transplant equine ATG (ATGAM, Pharmacia and Upjohn, Kalamazoo, MI, USA) or rabbit ATG (Thymoglobulin, Sanofi, Bridgewater, NJ, USA), and DBMT on day 0 followed by a one month course of calcineurin inhibitor. Recipients that underwent transplants after 2002 were also treated with co-stimulatory blockade either with anti-CD154 mAb (American Type Culture Collection) or belatacept (NULOJIX, Bristol-Myers Squibb Company, Princeton, NJ, USA). When DBMT was performed after KTx, the kidney transplant recipients were treated with a conventional immunosuppressive regimen consisting of tacrolimus (TAC, Astellas Pharma, Osaka, Japan), mycophenolate mofetil (MMF, Roche Inc., Nutley, NJ, USA), and methylprednisolone (MP) until DBMT.

Figure 1.

The nonmyeloablative conditioning regimen and results of transplants. (A) The pre-DBMT conditioning regimen includes low dose TBI (1.5 Gy × 2 on days −6 and −5), TI (7 Gy on day −1), anti-T cell antibodies (thymocyte globulin with or without anti-CD8 mAb). The recipients underwent DBMT either simultaneously or 4 months after kidney transplantation (KTx), followed by costimulatory blockade (anti-CD154 mAb or belatacept) and a 28-day course of cyclosporine, after which no immunosuppression was given.⁹ (B) Tolerant recipients (TOL) achieved long-term kidney allograft survival with stable kidney function (Serum creatinine: 1.0 - 2.0 mg/dl) without evidence of rejection. (C) Recipients with chronic antibody-mediated rejection (CAMR) initially did well but lost kidney function after developing CAMR with DSA. (D) Recipients that developed T cell mediated rejection (TCMR) failed to survive long-term with rapid rise of serum creatinine. (E) Renal allograft survival rates in TOL recipients were significantly longer than those in CAMR (p=0.0002) or TCMR (p<0.0001) recipients by Gehan-Breslow-Wilcoxon test.

Renal allograft FFPE samples

Protocol kidney allograft biopsies were taken approximately every three months during the first year. Indication biopsies were taken for any episode of unexplained renal allograft dysfunction. All biopsies were processed for routine light microscopy and immunohistochemistry (including Foxp3 and C4d stains) and diagnosed and scored according to Banff 2013 criteria by three pathologists (I.A.R., R.N.S. and R.B.C.). In the current study, there were 256 FFPE renal allograft samples from recipients that achieved tolerance (TOL, n=13), chronic antibody mediated rejection (CAMR, n=13), or T cell mediated rejection (TCMR, n=15). Tolerance was defined as the absence of DSA and C4d deposition with no histological evidence of CAMR for more than one year without maintenance immunosuppression. Since spontaneous resolution of subclinical TCMR were often observed, recipients that had a history of subclinical TCMR were not excluded from TOL. The MHC disparity and results of transplant in each recipient are summarized in Supplemental Table 1.

FFPE RNA isolation

Three consecutive 20 μm curls were obtained from each FFPE block, with microtome blade replacement and equipment sterilization with RNase AWAY (Life Technologies, Carlsbad, CA, USA) between blocks. Curls were immediately transferred to sterile microcentrifuge tubes and stored at room temperature. Xylene deparaffinization and RNA extraction were performed with the Recover All Total Nucleic Acid Isolation Kit for FFPE (Life Technologies, Carlsbad, CA, USA). RNA concentration and purity were measured with a Nano-Drop 2000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA).

NanoString gene expression analysis

The gene set included 52 oligonucleotide probes specific to Macaca fascicularis sequences manufactured by Integrated DNA Technologies (Coralville, IA, USA). These genes were chosen as they have been reported to be relevant to basic transplant immune responses, rejection and tolerance. ^{18–20,25–29,12,22–24,30,30,34,35} Four housekeeping genes (ACTB, GAPDH, HPRT1, and LDHA) were also included in the gene set for biological normalization purposes. Gene expression was quantified in FFPE-derived RNA isolates using the nCounter Elements assay (NanoString Technologies, Seattle, WA, USA). Quality control assessment and normalization of raw NanoString gene expression counts were performed with nSolver Analysis Software version 3.0 (NanoString Technologies, Seattle, WA, USA). The manufacturer’s recommended default parameters for quality control flagging were used for imaging (field of view registration <75%), binding density (<0.05 or >2.25), positive control linearity (R² value < 0.95), and positive control limit of detection. Background subtraction was performed for each sample by subtracting the mean of the negative controls from all data points. Each sample was first normalized to the geometric mean of the positive controls (with default flagging of normalization factors <0.3 and >3), followed by normalization to the geometric mean of the housekeeping genes (with default flagging of normalization factors <0.1 and >10). All raw data are shown in the supplemental Table 3.

Statistical analysis

In the present study, all previous samples were analyzed without randomization. Sample sizes were not based on power calculations. No data were excluded from the analyses. The code-blinded samples were sent to the University of Alberta, where the assays were performed in a blinded fashion. Post-normalization statistical analyses were performed with log-transformed mRNA expression levels. Between-group mean expression levels were compared by using longitudinal linear mixed effects model by which the time dependent piecewise (i.e., within- and after post-DBMT) group comparisons were conducted by pair-wise contrasts of the three groups (Proc Mixed, SAS 9.4, SAS Institute Inc., Cary, NC, USA.). Since all genes tested in this study have previously been reported to be relevant to basic transplant immune responses, rejection and tolerance, P-values were unadjusted for multiple comparisons, and values less than 0.05 were considered statistically significant. Receiver operating characteristic (ROC) curve analyses were performed using GraphPad Prism7 (GraphPad Software, La Jolla, CA, USA). Gene expression heat maps were created using hierarchical clustering by Euclidean distance with R version 3.4.3 (R Foundation for Statistical Computing, Vienna, Austria). Renal allograft survival was compared by Gehan-Breslow-Wilcoxon test using GraphPad Prism7 (GraphPad Software, La Jolla, CA, USA).

RESULTS

Clinical courses and pathological features of recipients in TOL, CAMR, and TCMR

As reported previously,^5–11 all recipients received major histocompatibility complex (MHC) mismatched kidney transplantation (KTx) and donor bone marrow transplantation (DBMT) following various non-myeloablative conditioning regimens, including total body irradiation (TBI) and thymic irradiation (TI), anti-T cell antibodies, costimulatory blockade, and a one month course of cyclosporine after which no immunosuppression was administered. DBMT was performed either immediately or several months after KTx (Figure 1A). Recipients developed various levels of transient mixed chimerism detectable for 2 - 8 weeks after DBMT. The recipients were categorized into three groups, tolerance (TOL), chronic antibody-mediated rejection (CAMR), and T cell mediated rejection (TCMR), depending upon the eventual outcome (Table 1). There was no significant difference in recipient and donor MHC disparity amongst the three groups (Supplemental Table 1) but recipients that developed higher maximum levels (>3.1%) of lymphoid chimerism showed a higher chance of achieving renal allograft tolerance.³⁶

Table 1.

Groups and Results

Group	N	Chimerism	DSA	C4d	GST (days)
TOL	13	13/13	0/13	0/13	5983, 4328, 3464, 2497, 1340,1021, 857, 834, 796, 758, 728, 468, 378
CAMR	13	12/13	12/13	11/11	2023, 1825, 1260, 1143, 843, 837, 791, 761, 703, 663, 434, 258, 148
TCMR	15	9/15	9/15	6/15	268,246, 217, 158, 141, 135, 135, 125, 116, 89, 72,71,69,64,58

In the TOL group, 13 recipients survived for more than one year with normal renal allograft function (Figure 1B) and renal allograft survival rates in TOL recipients appeared significantly longer than those in CAMR (p=0.002) or TCMR recipients (p=0.0003) (Figure 1E)(all raw data were also shown in the supplemental Table 3). They have never developed evidence of donor specific antibody (DSA) or any form of irreversible rejection (Figures 2A, 2B and 2C) during the entire observation period (1 - 16 years, mean observation time 1840 ± 1724 days). These recipients were often noted to have FOXP3⁺ cell-rich lymphoid aggregates focally around arteries and glomeruli (Figures 2B and 2C).³⁷

Figure 2.

Histopathology of renal allografts. (A-C): Renal allograft biopsies taken from a representative tolerant recipient. (A) Normal renal allograft biopsy taken from TOL on day 126 post-DBMT. (B) The biopsy taken on day 796 post-DBMT showed no rejection but a Treg-rich lymphoid aggregate around glomeruli and artery. (C) Immunohistochemical staining showed prominent FOXP3⁺ cells (stained brown) in the aggregate. (D-F): Renal allograft biopsies taken from a CAMR recipient. (D) Normal renal allograft biopsy on day 105 post-DBMT. (E) However, this recipient developed DSA after one year and allograft biopsy taken on day 663 post-DBMT showed transplant glomerulopathy with duplication of glomerular basement membrane and (F) prominent peritubular capillary C4d deposition. (G-I): Renal allograft biopsies taken from a recipient with TCMR. (G) Despite normal allograft function, borderline change was observed on day 54 post-DBMT. (H) This recipient then developed rapid deterioration of renal function on day 116 with marked mononuclear cell infiltration and endothelial injury. (I) Immunohistochemical stain of day 116 sample showed endothelialitis with CD3⁺ cell infiltration.

In the CAMR group, 13 recipients that initially did well (Figures 1C and 2D) eventually developed DSA and CAMR (Figure 2E) with prominent peritubular capillary C4d deposition (Figure 2F) and endothelial cell proliferation. They were euthanized at 899 ± 550 days because of progressively compromised renal allograft function (Figure 1C and 1E).

In the TCMR group, 15 recipients failed to develop multilineage mixed chimerism and lost renal allograft function due to TCMR soon after the discontinuation of immunosuppression (mean graft survival time 132 ± 69 days) (Figures 1D, 1E, 2G, 2H and 2I).

FOXP3 and Treg-related gene signatures were significantly higher in TOL

To investigate the kinetics of the intra-renal gene signatures in these recipients, we analyzed mRNA expression in biopsy or autopsy formalin-fixed and paraffin-embedded (FFPE) samples taken at various post-transplant time points. Using the NanoString nCounter platform, sufficient RNA (mean 138.5 ng/μL) was isolated from 242 of the 256 (94.5%) FFPE renal allograft samples. Those RNAs were tested for expression of 52 RNA genes that have been reported to be significantly associated with tolerance or rejection (Table 2). The transcript relative expression levels in early (within 1 year) (Figure 3) post-DBMT are shown on the heat map and statistical analyses of each gene are summarized in Supplemental Table 2. Since all recipients with TCMR did not survive beyond 268 days, the late analyses include only TOL and CAMR recipients.

Table 2.

52 mRNAs tested

Gene symbol	Gene name	Association
CD3D	CD3 delta chain	T cell
CD4	CD4	T cell
CD8A	CD8 alpha chain	T cell
TBX21	T-box 21 (T-bet)	T cell
GATA3	GATA binding protein 3	T cell
RORγt	RAR-related orphan receptor gamma t/2	T cell
FOXP3	Forkhead box P3	T cell
ICOS	Inducible T-cell costimulator (CD278)	T cell
IL6R	Interleukin 6 receptor (CD126)	T cell
EPO	Erythropoietin	T cell34
MS4A1	Membrane-spanning 4-domains, subfamily A, member 1 (CD20)	B cell23,24,58
FCGR3A	Fc fragment of IgG, low affinity IIIa, receptor (CD16a)	NK cell27,29
FGFBP2	Fibroblast growth factor binding protein 2 (KSP37)	NK cell25,29
KLRF1	Killer cell lectin-like receptor F1 (NKp80)	NK cell19,20
MYBL1	V-myb avian myeloblastosis viral oncogene homolog-like 1	NK cell19
SH2D1B	SH2 domain containing 1B (EAT2)	NK cell19,24
GNLY	Granulysin	NK cells
GZMB	Granzyme B (granzyme 2, CTL-associated serine esterase 1)	NK/T cells
BCL2	B-cell lymphoma 2	T/B/NK cells58
IFNG	Interferon, gamma	Cytokine
TNF	Tumor necrosis factor	Cytokine
TGFB1	Transforming growth factor, beta 1	Cytokine
IL21	Interleukin 21	Cytokine
IL2	Interleukin 2	Cytokine
IL17	Interleukin 17	Cytokine
IL10	Interleukin 10	Cytokine31
IL1RL1	Interleukin 1 receptor-like 1 (IL33R)	Cytokine
IL4	Interleukin 4	Cytokine
CX3CR1	CX3C chemokine receptor 1 (fractalkine receptor)	Cytokine19,25
CAV1	Caveolin 1	Endothelium19,20
CD34	CD34 antigen	Endothelium19,20
CDH13	Cadherin 13	Endothelium19,25
CDH5	Cadherin 5	Endothelium19,20
DARC	Duffy blood group, atypical chemokine receptor	Endothelium19,20
KLF4	Kruppel-like factor 4	Endothelium19,20
MALL	Mal, T-cell differentiation protein-like	Endothelium19,20
PALMD	Palmdelphin	Endothelium19,20
PECAM1	Platelet/endothelial cell adhesion molecule 1 (CD31)	Endothelium19,20
PLA1A	Phospholipase A1 member A	Endothelium19,20
PLAT	Plasminogen activator, tissue	Endothelium19,20
PSMB10	Proteasome subunit, beta type, 10	Endothelium19,20
RHOJ	Ras homolog family member J	Endothelium19,20
ROBO4	Roundabout, axon guidance receptor, homolog 4	Endothelium19,20
SELE	Selectin E	Endothelium19,20
SOX7	Sex determining region Y-box 7	Endothelium19,20
TEK	Tyrosine kinase, endothelial	Endothelium19,20
THBD	Thrombomodulin	Endothelium19,20
VWF	Von Willebrand factor	Endothelium19,20
RPS6KB1	Ribosomal protein S6 kinase beta 1	Endothelium19,20
RPS6	Ribosomal protein S6	Endothelium19,20
TRIB1	Tribbles pseudokinase 1	Endothelium19,20,40,59
CD74	Major histocompatibility complex, class II invariant chain	Endothelium19,20,60

Figure 3.

The transcript relative expression levels in early (within 1 year) post-DBMT. The transcript relative expression levels are shown on the heat map, in which red values indicate over-expression and green values indicate under-expression.

The most prominent difference observed among the three groups was FOXP3, which was significantly higher in TOL for several years, after which values slowly declined to baseline levels. FOXP3 was significantly higher in TOL than in both CAMR (p<0.01) and TCMR (p<0.001) early (within 1 year) post-DBMT and remained higher than CAMR (p<0.05) after 1 year. Although GATA3 and RORγt were significantly higher in TOL and CAMR (p<0.001), when compared with TCMR no difference was observed between TOL and CAMR within 1 year. GATA3 then became significantly higher in TOL than in CAMR (p<0.01) after 1 year. In addition, IL10 and TGFB, which have been reported to be important for Treg function/expansion, appeared weakly higher (p<0.05) in TOL. Conversely, IL6R and TRIB1, which have been reported to suppress Treg generarion^38,39 or high in chronic rejection⁴⁰, were higher in TOL than in CAMR (Figures 3 and 4A and Supplemental Table 2).

Figure 4.

Long-term kinetics of mRNA expression in the renal allografts. (A) FOXP3 was significantly higher in TOL for several years, after which it slowly declined to baseline levels. It was significantly higher in TOL than in both CAMR (p<0.01) and TCMR (p<0.001) within 1-year post-DBMT and remained higher than CAMR (p<0.05) after 1-year post-DBMT. GATA3 and RORγt were significantly higher in TOL and CAMR (p<0.001) compared with TCMR. No difference was observed between TOL and CAMR within 1-year post-DBMT. GATA3 became significantly higher in TOL than in CAMR (p<0.01) after 1 year. IL10 and TGFB, which have been reported to be important for Treg function/expansion, were weakly higher (p<0.05) in TOL. Conversely, IL6R and TRIB1, which have been negatively reported for Treg expansion or tolerance, were also high in TOL. (B) Gene signatures of inflammatory cytokines, such as IFNG, GZMB, IL1RL1 and IL4 were also significantly high in TCMR. (C) In CAMR, CAV1 was significantly higher than TOL both within and after 1 year, while VWF, MALL and PECAM1 became significantly higher than TOL only after 1 year.

*p<0.05, **p<0.01, ***p<0.001, ****p<0.0001, Green line, red line and blue line indicates the mRNA values in TOL, CAMR and TCMR, respectively.

mRNAs representing inflammatory cytokines are lowest in TOL

As anticipated from the histological findings with severe T cell infiltration, expression of CD4 was highest in TCMR (Supplemental Table 2). Gene signatures of inflammatory cytokines, such as IFNG, GZMB, GNLY, FCGR3A, IL1RL1 and IL4 were also significantly high in TCMR (Figures 3A and 4B).

Gene signatures of endothelial activation are a prominent feature in CAMR

In CAMR, although histological abnormalities were typically not observed within 1-year post-DBMT, endothelium-associated transcripts, such as CAV, was already significantly higher than in TOL both within and after 1 year, while MALL, DARC, VWF and PECAM1 became significantly higher than in TOL only after 1 year (Figures 3 and 4C). These results suggest that endothelial activation is the most prominent feature discriminating TOL from CAMR even before histological changes become evident.

Intra-graft mRNA profiles reliably distinguish TOL from CAMR

Since it is clinically important to predict transplant outcome before irreversible immunological responses take place, we performed receiver operating characteristic (ROC) curve analysis to identify which gene signature detected in the renal allografts before 1-year post-DBMT, when no histologic features of CAMR were evident, most reliably predicted transplant outcome. This analysis revealed that FOXP3 levels reliably differentiated TOL from CAMR (Figure 5A). In contrast, CAV1 (Figure 5B) reliably discriminated CAMR from TOL.

Figure 5.

The ROC analyses of mRNAs early after transplant (<1 year). (A) FOXP3 levels reliably differentiated TOL from CAMR (p<0.01). Conversely, CAV1 (B) reliably discriminated CAMR from TOL (p<0.05).

DISCUSSION

To further elucidate the mechanism by which the renal allograft itself plays a critical role in induction/maintenance of tolerance, we retrospectively analyzed 52 mRNA signatures in over 250 routine biopsy and autopsy FFPE samples from CKBMT recipients archived over the last 25 years. These genes were selected based on previously published reports that have identified these particular genes to be significantly associated with rejection or tolerance in KTx recipients.^{12,19,20,22–31} Although these recipients received slightly different conditioning regimens (e.g. anti-CD154 mAb vs. belatacept), all treatments were completed by 4 weeks post-Tx, after which no immunosuppressive drugs were administered. Since biopsy samples included in this study were taken after 2-3 months post Tx, we considered that different treatments administered during peri-transplant periods did not directly affect mRNA expression in the biopsy samples.

First, we established the dynamic time-dependent kinetics of intragraft mRNA expression. We observed that some genes which are significantly elevated in TOL in the first couple of years slowly declined over the following 3 - 10 years. Therefore, to perform our statistical analysis, we separated the data into two time periods: early (<1 year) and late (≥1 year) post-DBMT. The decision to separate the data at 1 year after transplant was based on our clinical observation that histological evidence of CAMR typically did not become evident until 1-year post-DBMT. If the insidious immune responses of CAMR could be detected before one year, reinstitution of immunosuppression might limit future development of CAMR.

The most prominent difference observed among all three groups was FOXP3, which was significantly higher in TOL than in CAMR both early and late after transplantation. We also identified significantly higher expression of GATA3 in TOL. Although GATA3 is a transcriptional factor for Th2, the critical importance of GATA3 co-expression for Treg function has recently been reported. ^41,42 The relatively high expression of RORγt in TOL may also be explained by the recent report of Yang et al., which demonstrated the importance of RORγt co-expression as a stable Treg lineage with enhanced suppressive capacity.⁴³ IL10 and TGFB have been shown to be critically important for Treg function and expansion, ^15,35,44,45 which is also consistent with local enrichment of Tregs in the graft. On the other hand, higher TRIB1 in TOL may contradict the report by Ashton-Chess J. et al., which showed significantly high intrarenal TRIB1 in CAMR.⁴⁰ They suggested that TRIB1 and FOXP3 interact together in Tregs and overexpression of TRIB1 was associated with a decrease in Treg proliferation.³⁹ Significantly higher TRIB1 in TOL in our study may suggest that overexpression of TRIB1 in Tregs may not necessarily inhibit Treg expansion in the graft if other factors, such as co-expression of GATA3, RORγt, or TGFB, are present. Significantly high IL6R in the current study may also contradict enriched Tregs,³⁸ but IL6-IL6R upregulation may be important to induce RORγt expression ⁴⁶ on Tregs to enhance their suppressive capacity.⁴³

Histologic observations of kidneys in the TOL group of our current study show lymphoid aggregates with abundant Tregs, similar to those that have been observed in a mouse spontaneous kidney allograft tolerance model. Miyajima et al. called this lymphoid aggregate Treg-rich organized lymphoid structure (TOLS) and showed that deletion of intra-graft Tregs abrogated renal allograft tolerance.³⁷ Using the same mouse model, Hu et al. also found expanded Tregs in the kidney and in draining lymph nodes that expressed elevated levels of TGF-β, IL-10, and IFNγ. They also showed abrogation of tolerance by deleting these Tregs.⁴⁷ The importance of intragraft Tregs has also been demonstrated by Cobbold et al. in a mouse skin tolerance model induced by non-depleting anti-CD4 mAb. High levels of FOXP3 mRNA and glucocorticoid-induced TNFR superfamily member 18 (GITR)⁺CD25⁺ were found within the skin grafts of long-term tolerant recipients. Since splenic T cells from tolerant mice responded normally to the donor antigen, intra-graft Tregs were postulated to protect the skin graft from rejection by recipient splenic T cells.⁴⁸ Furthermore, surgically removed Treg-rich skin grafts from the tolerant recipients were successfully retransplanted onto the same strain recipient without additional anti-CD4 mAb treatment. This skin graft could be rejected by ablating Tregs from the skin graft, indicating that intra-graft Tregs play a critical role in infectious tolerance.⁴⁹

Enriched intra-graft Tregs may be expanded tTregs or pTregs converted from non-Tregs, but it has been difficult to differentiate these two populations due to a lack of easily used molecular markers to distinguish them. However, based on our previous studies which showed donor-specific Treg expansion converted from non-Tregs in the peripheral blood, ¹⁵ the enriched intra-graft Tregs may also be pTregs converted from non-Tregs.

The reason why initially high Foxp3⁺ Tregs in the renal allograft declined to baseline levels after several years while tolerance of the kidney persisted remains to be defined. We speculate that peripheral mechanisms of tolerance, although important initially, may be converted to deletional tolerance over time, as has been postulated in human recipients who achieve tolerance via the mixed chimerism approach.^50,51

Since B cells are not a major cell population identified in the renal allograft, MS41A (CD20), a trans-membrane protein expressed on pre- and mature B cells, was the only B cell- related mRNA included in this study. Peripheral blood studies of spontaneous tolerance in kidney recipients, who had discontinued their immunosuppression,^22–24 concluded that B cell related gene signatures and the relative expansion of the transitional B cell subsets appeared to be significant.^52,53 However, increased intrarenal expression of MS4A1 in TOL recipients was not observed in the current study. Although more studies with B cell related genes could be revealing, biomarkers identified in the peripheral blood of recipients achieving spontaneous tolerance after prolonged conventional immunosuppression may not correlate with local immune responses in the allograft following DBMT. Gene signatures of intentionally induced tolerance may be different from spontaneously induced tolerance.

Our studies revealed gene signatures of inflammatory responses, such as IFNG, GZMB, GNLY, FCGR3A (CD16a), IL1RL1, and IL4, to be high exclusively in TCMR as expected. Despite reports of increased expression of TBX21 (T-bet) in acute rejection or CAMR,^54,55 there was no significant difference in TBX21 among the groups in this study. This may also be due to the fact that TBX21 is not strictly specific to Th1 but can be expressed by other T cell subsets, such as Tregs.⁵⁶

As we have previously reported, the specific gene sets that represent chronic rejection, namely, CAV, VWF, and DARC, were exclusively high in CAMR.⁵⁷ These genes are all endothelium-related transcripts with long-standing associations with CAMR in humans. Their biological functions, which include endothelial injury, repair, activation, and angiogenesis, were upregulated even when histological changes of the endothelium were not clearly observed. Although histological changes in CAMR recipients were often not evident until one year after transplant, CAV1 was already significantly high in CAMR recipients. This emphasizes that FOXP3 and CAV1 should be useful to predict outcomes early after withdrawal of immunosuppression before irreversible destructive immune responses develop. From the ROC curve analysis, we found intra-graft FOXP3 reliably differentiates recipients that subsequently maintain tolerance from recipients that later develop CAMR (AUC: 0.82). On the other hand, since endothelium-associated transcripts such as CAV1 were significantly higher in CAMR compared with TOL, the ROC curve analysis also shows better AUC which discriminated those recipients destined to develop CAMR (AUC: 0.73).

In conclusion, the current study revealed that intra-graft enrichment of Foxp3⁺Tregs may be an important mechanism of renal allograft tolerance induced by transient mixed chimerism. Gene signatures specific to tolerance, especially FOXP3, can be useful biomarkers to reliably predict the long-term results of the allograft after induction of tolerance via this approach. These biomarkers may be more broadly applicable to humans and to other protocols and will be the subject of future clinical trials for tolerance. The application of new approaches, such as RNAseq with more expanded gene sets may unravel new mechanisms of tolerance.

Abbreviations:

ATG

anti-thymocyte globulin

CAMR

chronic antibody-mediated rejection

CNI

calcineurin inhibitor

CKBMT

combined kidney and bone marrow transplantation

BMT

bone marrow transplantation

DSA

donor-specific antibody

FFPE

formalin-fixed and paraffin-embedded

FOXP3

forkhead box P3

KTx

kidney transplantation

NHPs

non-human primates

NDAR

no diagnostic abnormality

RNA

ribonucleic acid

TBI

total body irradiation

TCMR

T cell mediated rejection

TI

thymic irradiation

TOL

tolerance

Tregs

regulatory T cells