Variations in DNA methylation of interferon gamma and programmed death 1 in allograft rejection after kidney transplantation

The role of DNA methylation in the regulation of the anti-donor-directed immune response after organ transplantation is unknown. Here, we studied the methylation of two mediators of the immune response: the pro-inflammatory cytokine interferon γ (IFNγ) and the inhibitory receptor programmed death 1 (PD1) in naïve and memory CD8+ T cell subsets in kidney transplant recipients receiving immunosuppressive medication. Both recipients experiencing an episode of acute allograft rejection (rejectors) as well as recipients without rejection (non-rejectors) were included. CpGs in the promoter regions of both IFNγ and PD1 were significantly (p < 0.001) higher methylated in the naïve CD8+ T cells compared to the memory T cell subsets. The methylation status of both IFNγ and PD1 inversely correlated with the percentage of IFNγ or PD1-producing cells. Before transplantation, the methylation status of both IFNγ and PD1 was not significantly different from healthy donors. At 3 months after transplantation, irrespective of rejection and subsequent anti-rejection therapy, the IFNy methylation was significantly higher in the differentiated effector memory CD45RA+ (EMRA) CD8+ T cells (p = 0.01) whereas the PD1 methylation was significantly higher in all memory CD8+ T cell subsets (CD27+ memory; p = 0.02: CD27− memory; p = 0.02: EMRA; p = 0.002). Comparing the increase in methylation in the first 3 months after transplantation between rejectors and non-rejectors demonstrated a significantly more prominent increase in the PD1 methylation in the CD27− memory CD8+ T cells in rejectors (increase in rejectors 14%, increase in non-rejectors 1.9%, p = 0.04). The increase in DNA methylation in the other memory CD8+ T cells was not significantly different between rejectors and non-rejectors. At 12 months after transplantation, the methylation of both IFNγ and PD1 returned to baseline levels. The DNA methylation of both IFNγ and PD1 increases the first 3 months after transplantation in memory CD8+ T cells in kidney transplant recipients. This increase was irrespective of a rejection episode indicating that general factors of the kidney transplantation procedure, including the use of immunosuppressive medication, contribute to these variations in DNA methylation.


Background
Kidney transplantation is currently the best treatment option for patients with irreversible, end-stage kidney disease [1]. Successful kidney transplantation is hampered by different complications including immune-mediated complications such as acute rejection [2]. Several noninvasive biomarkers for acute rejection have been studied, including proteins involved in cytotoxic lymphocyte function (e.g., perforin and granzyme B), cytokines (e.g., interferon (IFN) γ), and immune-related chemokines (e.g., CXCL9 and CXCL10) [3,4]. Nevertheless, it remains difficult to predict and regulate the host immune response after transplantation. The host immune response is orchestrated by a tightly regulated cascade of gene expression changes which are regulated by epigenetic mechanisms like histone modifications, DNA methylation, microRNA interactions, and chromatin remodeling complexes [5][6][7][8]. Variations in these epigenetic mechanisms might serve as an additional marker to monitor the host immune response after organ transplantation.
An important player of the host immune response is the pro-inflammatory cytokine IFNγ, and high expression of IFNγ is associated with both acute and chronic allograft rejection [9][10][11]. The expression of IFNγ is regulated by DNA methylation with the addition of methyl groups on cytosine-phosphate-guanine sites (CpGs) in the IFNγ promoter region silencing its expression. The CpG methylation pattern of IFNy discriminates different T cell subsets. First, naïve (antigen unexperienced) T cells versus memory (antigen experienced) T cells (both CD4+ and CD8+ T cells) with memory T cells having a lower methylation profile [12][13][14]. Second, the different T helper cell (Th) subsets with Th1 cells being hypomethylated compared to the Th2 and Th17 subsets [15][16][17]. Another important molecule involved in the regulation of the antidonor immune response is the inhibitory receptor programmed cell death (PD) 1. Aggressive recipient T cells that attack the transplanted organ, the so-called alloreactive T cells, are inhibited by PD1 signaling. In addition, PD1 signaling promotes the generation of induced regulatory T cells [18,19]. The expression of PD1 is also dependent on DNA methylation and while mainly methylated in naïve T cells, PD1 is demethylated during differentiation into memory T cells [20].
Regulation of gene expression by DNA methylation is a well-known epigenetic mechanism with a critical role in physiological development and normal cell function by coordinating the lineage-and tissue-specific expression of genes [21]. DNA methylation is dynamic and susceptible to stimuli from the environment including internal stimuli like cytokines and hormones and external stimuli like chemical agents, pollutants, dietary components, and chronic viral infections [16,[22][23][24]. Aberrant DNA methylation profiles are associated with the pathogenesis of disease. Initially, DNA methylation was associated with tumor formation and progression [25], but later on, variations in DNA methylation have been associated with other diseases [26,27] including chronic kidney disease (CKD) [28,29] and immunemediated diseases such as rheumatoid arthritis [30] and allergy [31,32]. In addition, variations in DNA methylation of immune-related genes orchestrate the host immune response after organ transplantation [5][6][7][8].
Graft-infiltrating cytotoxic CD8+ T cells play a major role in the rejection process and elevated numbers of effector, and memory CD8+ T cell subsets are associated with an increased risk for acute rejection [33][34][35]. Here, we examined the influence of variations in DNA methylation of IFNγ and PD1 in different CD8+ T cell subsets on allograft rejection. The DNA methylation of IFNγ and PD1 was determined in kidney transplant recipients before and 3 and 12 months after transplantation, and both kidney transplant recipients who experienced a rejection episode within the first 3 months after transplantation and recipients who remained free from rejection were included. To exclude gender- [32] or chronic viral infection [24]-related differences, we first analyzed whether the DNA methylation of either IFNγ or PD1 was different in males versus females or in cytomegalovirus (CMV) seropositive healthy donors versus CMV seronegative healthy donors.

IFNγ methylation is significantly decreased in CMV seropositive individuals
In PBMCs of CMV seronegative healthy donors, the DNA methylation of IFNy was 51.2 ± 4.4% (mean ± SD). The IFNy methylation was significantly lower in PBMCs of age-matched CMV seropositive healthy kidney donors (45.1 ± 7.2%, p = 0.009; Fig. 1a). In both males and females, the methylation of IFNy was lower in the CMV seropositive individuals (Fig. 1a), and there was no significant difference between males and females. The DNA methylation of PD1 in PBMCs of CMV seronegative healthy donors was comparable to the PD1 methylation in CMV seropositive healthy donors (40.5 ± 5.3 versus 38.9 ± 6.3%; Fig. 1b). Subdividing the PBMCs into the different CD8+ T cell subsets (Fig. 1c) demonstrated significantly lower methylation of IFNγ in naïve, CD27+ memory, and CD27− memory CD8+ T cells in CMV seropositive individuals compared to CMV seronegative individuals (Fig. 1d). The methylation of PD1 was not significantly different between the CMV seropositive individuals and CMV seronegative individuals in all the studied CD8+ T cell subsets (Fig. 1e).

DNA methylation inversely correlates with protein expression
To determine whether variations in DNA methylation at the described CpGs [20,36] are associated with changes in protein expression, we measured the expression of IFNγ and PD1 in the different CD8+ T cell subsets (Fig. 2a). A clear-cut difference was observed between the naïve CD8+ T cells compared to the memory CD8+ T cells where 14.6 ± 16.4% (mean ± SD) of naïve CD8+ T cells expressed IFNγ versus 50.3 ± 18.9% of the CD27+ memory, 52.6 ± 20.6% of the CD27− memory and 66.1 ± 19.8% of the EMRA CD8+ T cells expressed IFNγ (p < 0.0001; Fig. 2b). In parallel, a significantly lower percentage of naïve CD8+ T cells expressed PD1 compared to the memory CD8+ T cell subsets (naïve 27.3 ± 16.5%, CD27+ memory 67.9 ± 5.1%, CD27− memory 68.4 ± 12.2%, and EMRA 51.4 ± 20.1; p < 0.0001; Fig. 2e). The highest percentage of IFNγ-expressing cells was found within the EMRA CD8+ T cells while the CD27+ and CD27− memory CD8+ T cell subsets contained the highest percentages of PD1-expressing cells. The DNA methylation of both IFNγ and PD1 demonstrated the opposite pattern with the highest percentage of methylation in naïve CD8+ T cells. Naïve CD8+ T cells were methylated for 55.2 ± 18.3% at the IFNγ locus and for 43.1 ± 10.7% at the PD1 locus. This methylation was significantly higher (p < 0.0001 for both IFNγ and PD1) compared to the different memory CD8+ T cell subsets (Fig. 2c, f ). This inverse relation between the DNA methylation and protein expression confirms the regulatory capacity of the studied CpGs (Fig. 2d, g).

Variations in DNA methylation in kidney transplant recipients before transplantation
Before kidney transplantation, the methylation of IFNγ in CMV seronegative kidney recipients was comparable to the and CMV seropositive healthy donors (n = 15; gray bars) in PBMCs (mean ± SD) and stratified by gender (box and whiskers min to max). Gating strategy of the different CD8+ memory T cell subsets in (c). The percentage of DNA methylation of IFNγ (d) and of PD1 (e) in CMV seropositive (n = 5; open bars) and CMV seronegative healthy donors (n = 5; gray bars) in cell sorted CD8+ T cell subsets; naïve, CD27+ memory, CD27− memory, and differentiated effector memory CD45RA+ (EMRA). Box and whiskers (min to max); *p < 0.05 and **p < 0.01 methylation levels in CMV seronegative healthy donors for naïve, CD27+ memory, CD27− memory, and EMRA CD8+ T cells (Fig. 3a). The same pattern was seen for the methylation of PD1 (Fig. 3b). Subdividing the transplant recipients into the ones that went on to experience a rejection after transplantation, the rejectors and the non-rejectors, did not reveal any significant differences in methylation of IFNγ nor PD1, either between the two recipient groups or in comparison to the healthy donors (data not shown).

Variations in DNA methylation in kidney transplant recipients after transplantation
After kidney transplantation, the percentage of methylation of IFNγ did not change significantly in the naïve, Fig. 2 IFNγ and PD1 protein expression and IFNγ and PD1 DNA methylation. FACS plots of IFNγ and PD1 expression in naïve, CD27+ memory, CD27− memory, and differentiated effector memory CD45RA+ (EMRA) CD8+ T cells in (a; representative example). Mean protein expression and percentage of DNA methylation in the different CD8+ T cell subsets in kidney transplant recipients before transplantation (n = 10; IFNγ in b-d and PD1 in e-g; mean ± SD). ***p < 0.001 CD27+ memory, and CD27− memory CD8+ T cells during the first year after transplantation ( Fig. 4a-c). In the EMRA CD8+ T cells, the methylation of IFNγ was significantly higher at 3 months after transplantation compared to the methylation before transplantation irrespective of rejection and the subsequent antirejection therapy (p = 0.01; Fig. 4d). Focusing on rejection demonstrated that the methylation of IFNγ was significantly higher at 3 months after transplantation in the rejectors (14.3% versus 6.3% before transplantation; p = 0.01) while the non-rejectors increased from 4.9 to 8.6% (not significant). Both rejectors and non-rejectors demonstrated elevated IFNγ methylation levels in the EMRA CD8+ T cells at 3 months after transplantation, but this increase in methylation was not significantly different between rejectors and non-rejectors (p = 0.3). At 1 year after transplantation, the methylation of IFNγ was comparable to the levels measured before transplantation.

Discussion
The clinical potential of DNA methylation in organ transplantation, either as diagnostic or prognostic biomarker or as therapeutic target has been proposed by many [5-8, 37, 38]. Nevertheless, this is the first study where DNA methylation of two selected genes, IFNγ and PD1, was actually studied in CD8+ T cells in a small cohort of human kidney transplant recipients over time in relation to acute allograft rejection. Irrespective of rejection, we observed at 3 months after transplantation significantly elevated DNA methylation levels of IFNγ in the differentiated EMRA CD8+ T cells, while the DNA methylation of PD1 was significantly higher in all CD8+ memory T cell subsets. This increase in IFNγ methylation was not significantly different between rejectors and non-rejectors, while the increase in PD1 methylation was significantly higher in the rejectors in the CD27− memory CD8+ T cells. In the other CD8+ memory T cells subsets (CD27+ memory and EMRA), the increase in DNA methylation of PD1 was not significantly different between rejectors and non-rejectors.
Kidney transplantation will activate the recipient's immune system accompanied by an increase in cytokine production, including production of the pro-inflammatory IFNγ [35,39,40], and upregulation of PD1 expression [41]. As protein expression inversely correlates with DNA methylation levels at gene promoter sites, kidney transplantation induces demethylation of genes involved in immune activation. However, for both IFNγ and PD1, an increase in DNA methylation was observed in rejectors and non-rejectors in the first 3 months after transplantation, indicative for lower expression levels of IFNγ and PD1. Likely, the expected demethylation is only detectable in the donor-antigen specific T cells. The low percentage of these cells within the selected CD8+ T cells explains why the expected decrease in methylation was not observed. The observed increase in IFNγ and PD1 DNA methylation most likely does not reflect the immune response against the foreign donor antigen but demonstrates a downregulation of the immune system achieved by the given immunosuppressive medication which nonspecifically block all T cell subsets. For example, the usage of prednisolone. In this study, prednisolone was tapered to 5 mg at month 3 and thereafter completely withdrawn. At 1 year after transplantation, the DNA methylation levels returned to baseline.
In a clinical transplantation setting, it is impossible to measure the DNA methylation of either IFNγ or PD1 just before rejection. Currently, rejection cannot be predicted as the moment of rejection strongly varies between individuals and therefore those samples are not available. Although the material was only available of a small number of patients, we had the unique opportunity to follow the same patients over time. Variations in DNA methylation are more profoundly found in the period after withdrawal of stress exposure (e.g., drugs) compared to the period during exposure [42,43]. Translation to the field of organ transplantation implies that after a rejection episode including anti-rejection therapy, rejectors would have more variations in DNA methylation compared to non-rejectors. However, this was not true for the methylation of either IFNγ or PD1 at 12 months after transplantation, indicating that allograft rejection has no imprinted effect on the DNA methylation of those immune genes.
Despite differences in immune activity of the distinct memory CD8+ T cell subsets, the variations in DNA methylation in either memory subset were comparable. The EMRA CD8+ T cells are potentially the most aggressive subtype with a strong cytolytic activity, while the CD27+ memory cells display weak cytolytic activity producing effector cytokines such as interleukin (IL) 2, IFNγ, tumor necrosis factor (TNF) α, and IL4 [44,45]. The CD27− memory CD8+ T cells, which are functionally in between the CD27+ memory CD8+ T cells and the EMRA CD8+ T cells, represent the smallest subpopulation and it is unclear why specifically these cells demonstrated a significant difference in increase in methylation of PD1 between rejectors and non-rejectors. DNA methylation is adjustable by cues from the environment, e.g., viral infections [20,24,46], though the exact cues and mechanisms remain largely unknown [16,22,23]. The uremic condition during chronic kidney disease (CKD) modifies DNA methylation profiles [47][48][49]. Although, before transplantation, we did not observe significant changes in the methylation of either PD1 or IFNγ compared to agematched healthy donors. Either the previously observed effect on DNA methylation is gene specific and not applicable to IFNy and PD1 or the included transplant recipients here had less severe kidney disease compared to the CKD patients studied previously.
In contrast to previous observations where males demonstrated a significantly higher DNA methylation of IFNγ compared to females [32], significant differences in DNA methylation between males and females were not observed. However, we observed a significantly lower % of IFNγ methylation in CMV seropositive healthy donors compared to CMV seronegative healthy donors. The effect of chronic CMV infection on DNA methylation is not documented yet, but the change of the composition of the T cell pool with a permanent increase in highly differentiated T cells with a more memory phenotype in CMV seropositive individuals [50] has been demonstrated repeatedly. Therefore, the lower % of IFNγ methylation in CMV seropositive individuals might be explained by the fact that memory T cells are less methylated at the IFNγ locus ( Fig. 2 and [12-14]). Nevertheless, also in selected CD8+ memory T cells, the methylation of IFNγ was significantly lower in the CMV seropositive individuals (Fig. 1), indicating that CMV infection not only affects the composition of the T cell compartment but also induces a more aggressive T cell phenotype since demethylation is associated with an increased IFNγ production.
Although we could not identify variations in DNA methylation of either IFNy or PD1 in CD8+ T cells which could either diagnose or predict allograft rejection after kidney transplantation, further research is needed to appreciate the clinical significance of variations in DNA methylation and other epigenetic mechanisms in kidney transplantation. Epigenetic biomarkers, mainly based on variations in DNA methylation, are well established in the diagnosis of cancer and are not only detectable in the affected tissue as well as in the urine or the peripheral blood [51,52]. Currently, the application of epigenetic biomarkers is extended to other complex diseases such as autoimmune diseases [30,53,54]. The increasing knowledge on the epigenetic regulation of immune cells will contribute to our understanding of the epigenetic regulation of the complex anti-donor immune response after kidney transplantation. Epigenetic variations precede changes in protein expression and cell function and thereby represent an early indicator of clinical complications. Accordingly, a more comprehensive understanding of the epigenetic regulation of the antidonor immune response will learn whether variations in DNA methylation can serve as predictive, diagnostic, or prognostic markers. Moreover, since DNA methylation is influenced by environmental cues, it might serve as a target for therapeutic intervention.
A genome-wide approach instead of selected immunoregulatory genes are a good option for future research. Genome-wide analysis enables the identification of variations in DNA methylation in all promoter regions as well as other gene regions including intragenic and intergenic regions [47,55,56]. Since DNA methylation profiles are cell-type specific [57], selected cell subsets involved in the anti-donor immune response (e.g., CD4+ T cell subsets, B cells, and macrophages), or even better the donor-antigen specific cells, should be analyzed. Another interesting, though technically more challenging option, is to analyze variations in DNA methylation in graft-infiltrating T cells. As variations in DNA methylation occur specifically in donor-antigen specific cells which are more abundantly present in the graft compared to the circulation.

Conclusions
After kidney transplantation, the DNA methylation of the promoter of both IFNγ and PD1 increases in the first 3 months and returns to baseline at 1 year after transplantation irrespective of rejection. These variations do not reflect the anti-donor immune response but are more likely the result of the transplantation procedure and the use of immunosuppressive medication.

Study population
Prior to the selection of kidney transplant recipients, we first determined whether cytomegalovirus (CMV) infection modulates DNA methylation of either IFNγ or PD1. Peripheral blood mononuclear cells (PBMCs) of 15 CMV seropositive healthy donors (age 52 years, range 38-71; 5 males and 10 females) and 15 age-matched CMV seronegative healthy donors (age 52 years, range 44-59; 11 males and 4 females) were studied. Of these 30 healthy donors in total, we selected 5 CMV seropositive and 5 CMV seronegative age-matched individuals to study the methylation status in different CD8+ T cell subsets. Based on the significant decrease in DNA methylation of IFNγ in CMV seropositive healthy donors, we included only CMV seronegative kidney transplant recipients who received their first kidney from a living donor. The DNA methylation of both IFNγ and PD1 was examined in different CD8+ T cell subsets in 5 recipients who developed a biopsy-proven acute cellular rejection within the first 3 months after transplantation (rejectors; Table 1) and 5 age-matched recipients who remained free from rejection the first year after transplantation (non-rejectors) and was compared to 5 agematched healthy donors (age 54 years, range 44-59). The different CD8+ T cell subsets were analyzed at different time points; before transplantation and 3 and 12 months after transplantation. The selected CMV seronegative recipients all received a kidney from a CMV seronegative donor and received basiliximab as induction therapy. After transplantation, recipients received standard triple maintenance therapy consisting of prednisolone (tapered after 3 months), mycophenalate mofetil (MMF), and tacrolimus. Anti-rejection therapy consisted of methylprednisolone (1 g per day) on three consecutive days followed in some cases by anti-thymocyte globulin (ATG; n = 2) or alemtuzumab (n = 1).

Bisulfite conversion
PBMCs and the FACS-sorted CD8+ T cell subsets were digested with proteinase K and treated with bisulfite using the EZ DNA Methylation-Direct Kit (Zymo Research from Base Clear Lab products, Leiden, The Netherlands), according to the manufacturer's instructions. During bisulfite treatment, unmethylated cytosines were converted into uracil, whereas methylated cytosines remained unchanged.
PCR amplifications were performed with the Pyromark PCR Kit from Qiagen with each primer in a concentration of 0.2 μM. The PCR conditions were 15 min at 95°C , 45 cycles of 30 s 94°C, 30 s 58°C for IFNγ and 56°C for PD1 and 30 s 72°C followed by 10 min at 72°C and on hold at 21°C. After visualization of the appropriately sized PCR product on a 1% agarose gel, the PCR product was sequenced using a PyroMark Q24 pyrosequencer (Qiagen) with the following minor revisions to the manufacturer's instructions: to immobilize the PCR product, 1 μl Streptavidin Sepharose High Performance Beads (GE Healtcare) were used per sequence reaction and

Statistical analysis
To identify differences between groups, the unpaired t test, Mann-Whitney U test, and ANOVA were used as appropriate. To determine differences after kidney transplantation over time between rejectors and nonrejectors, we used multilevel analysis with the percentage of DNA methylation as outcome. Predictors were different individuals (rejectors and non-rejectors), time also as categorical predictor (levels 0 (before transplantation), 3 and 12 months after transplantation), and individuals as random intercept. Each model was applied for the four different cell types studied; naïve, CD27+ memory, CD27− memory, and EMRA CD8+ T cell subsets. Afterwards, we added models with interaction between type of individual and time. The first model describes the same pattern over time for both rejectors and non-rejectors while the second one enables to estimate and test different trends in time for rejectors and non-rejectors. The estimates and standard errors were transformed to CI's and p values. We used the package R version 3.1.2 and libraries lmer and lmerTest. A p value of <0.05 was considered statistically significant.