Increased Pre-Transplant Frequency of CD28⁺ CD4⁺ T_EM Predicts Belatacept-Resistant Rejection in Human Renal Transplant Recipients

Abstract

While most human T cells express the CD28 costimulatory molecule constitutively, it is well-known that age, inflammation, and viral infection can drive the generation of CD28^null T cells. In vitro studies have demonstrated that CD28^null cell effector function is not impacted by the presence of the CD28 costimulation blocker belatacept. As such, a prevailing hypothesis suggests that CD28^null cells may precipitate costimulation blockade-resistant rejection. However, CD28⁺ cells possess more proliferative and multi-functional capacity, factors that may increase their ability to successfully mediate rejection. Here, we performed a retrospective immunophenotypic analysis of adult renal transplant recipients who experienced acute rejection on belatacept treatment as compared to those that did not. Intriguingly, our findings suggest that patients possessing higher frequency of CD28⁺ CD4⁺ T_EM prior to transplant were more likely to experience acute rejection following treatment with a belatacept-based immunosuppressive regimen. Mechanistically, CD28⁺ CD4⁺ T_EM contained significantly more IL-2 producers. In contrast, CD28^null CD4⁺ T_EM isolated from stable belatacept-treated patients exhibited higher expression of the 2B4 coinhibitory molecule as compared to those isolated from patients who rejected. These data raise the possibility that pre-transplant frequencies of CD28⁺ CD4⁺ T_EM could be used as a biomarker to predict risk of rejection following treatment with belatacept.

Introduction

Current standard of care immunosuppression for kidney transplantation is focused on the use of calcineurin inhibitors (CNI), and while these reagents are efficacious in inhibiting allograft rejection in the vast majority of patients, they possess a number of off-target effects including nephrotoxicity, hypertension, diabetes, and infectious and malignant complications. Approved by the FDA in 2011, belatacept has emerged as an alternative to CNI for the prevention of kidney transplant rejection and offers improved side-effect profile by targeting CD28 signals important for optimal T cell activation during an alloimmune response. However, both early clinical trials and more recent real-world experience have demonstrated that a subset of patients on belatacept-based regimens experience acute cellular rejection in the first 6 months post-transplant (1, 2). Transplant practitioners are therefore faced with the challenge of assessing the risk-benefit ratio of these two regimens for individual patients without clinical tools or fundamental knowledge to predict (and possibly mitigate) rejection on belatacept-based regimens.

T cells express a myriad of inducible costimulatory and coinhibitory receptors during activation and differentiation, and these function to fine-tune the magnitude and character of an immune response (3–7). CD28 generates a critical costimulatory signal transduced into T cells following ligation of CD80 and/or CD86 on the surface of APCs (4). Most T cells express CD28 constitutively, however in the setting of advanced age or chronic inflammation, humans and NHP accumulate populations of both CD4⁺ and CD8⁺ CD28^null cells (8). Interestingly, emergence of these CD28^null CD4⁺ and CD8⁺ T cell populations has been described in the peripheral blood of patients with chronic kidney disease (CKD) (9, 10). Importantly, patients with kidney failure awaiting transplantation have been described to have variable accumulation of CD28^null CD4⁺ T cells in peripheral blood (11–14). Because belatacept functions to inhibit T cell activation by binding to CD80 and CD86 and thereby preventing CD28-mediated costimulation (15), it stands to reason that CD28^null cells would be resistant to the effects of belatacept. Indeed, in vitro studies have shown that CD8⁺ CD28^null cell effector function is not impacted by the addition of belatacept to cell cultures (16–18). Based on these data, the prevailing hypothesis in the transplant immunology community is that CD4⁺ and/or CD8⁺ CD28^null cells mediate costimulation blockade-resistant rejection (19–21), and that patients who reject their kidney graft during co-stimulation blockade may have increased pre-transplant frequencies of CD4⁺ and/or CD8⁺ CD28^null T cells (21). Despite this assumption, there are no published reports correlating pre-transplant CD4⁺ or CD8⁺ CD28^null frequency with risk of costimulation blockade-resistant rejection.

As such, these findings have raised an important and unanswered question; while CD28^null T cells may be costimulation-independent, are they actually capable of rejecting a graft? It is known that the CD28^null cells that emerge as a result of increasing age and/ or chronic immune activation often exhibit phenotypic and functional characteristics of immune senescence. For example, CD8⁺ CD28^null T cells demonstrate diminished proliferative capacity and oligoclonality of their TCR (22–24). Still, these cells possess enhanced immediate cytotoxic functions (22–24) and thus could potentially mediate graft rejection through direct cytolysis. There are also contradictory reports of the functional capacity of CD28^null cells in various disease states. CD4⁺CD28^null T cells isolated from patients on dialysis express higher basal levels of cytolytic molecules (14). Conversely, increased frequency of CD4⁺CD28^null cells in elderly kidney transplant recipients has been associated with improved graft survival under standard CNI immunosuppression, likely due to immunosenescence (12, 25). A more recent study similarly found that immunological ageing-related expansion of highly differentiated CD4⁺CD28^null T cells was associated with a lower risk of acute rejection (26). In contrast, increased frequency of CD8⁺CD28^null cells has been associated with increased rejection in young recipients on CNI therapy (25), and frequencies of either CD4⁺ or CD8⁺ CD28^null T cells failed to predict infectious complications following renal transplantation (27). Here, we performed a retrospective immunophenotypic analysis of adult renal transplant recipients who experienced acute rejection on belatacept treatment as compared to those that did not. Intriguingly, in contrast to the prevailing (yet untested) hypothesis in the field, our findings suggest that patients possessing higher frequency of CD28^null T cells within the CD4⁺ or CD8⁺ T_EM compartment prior to transplant were less likely to experience acute rejection following treatment with a belatacept-based immunosuppressive regimen.

Methods

Patient immunosuppression and sample collection

All patients undergoing renal transplantation at Emory University Hospital between the years 2009 and 2015 were enrolled in an immune monitoring protocol approved by Emory University’s Institutional Review Board (IRB #00046593) after informed consent was obtained. Patients who were Epstein Barr virus (EBV) seronegative, HIV⁺, had a history of post-transplant lymphoproliferative disease (PTLD), lymphoma or other hematologic malignancy, were undergoing treatment for latent tuberculosis or were the recipient of a simultaneous extra-renal organ were excluded from belatacept treatment. Patients treated with belatacept received intravenous infusion of belatacept (10mg/kg) during surgery and on post-transplant days 28, 56, and 84 with subsequent doses (5mg/kg) given every 4 weeks thereafter. Belatacept-treated patients also received anti-IL-2R induction (basiliximab 20mg iv on days 0 and 3 or 4), mycophenolate mofetil (MMF, 1g twice daily), a short steroid taper (methylprednisolone 500mg iv intra-operatively, 250 mg iv d1, 125mg iv d2, and prednisone 5mg d3 and daily thereafter), and a tacrolimus taper over the first 3–9 months (target trough levels 5–12ng/ml) (Adams et al., AJT 2016, In press). The cohort of patients treated with the standard tacrolimus-based regimen (Tac) also included anti-IL-2R induction (basiliximab 20mg iv on days 0 and 3 or 4), mycophenolate mofetil (MMF, 1g twice daily) and a short steroid taper (methylprednisolone 500mg iv intra-operatively, 250 mg iv d1, 125mg iv d2, and prednisone 5mg d3 and daily thereafter). Patient samples were acquired either prior to transplantation (baseline), or at a follow-up timepoint 1–5 months post transplant (for stables) or within 3 days of the diagnosis of acute rejection, before any anti-rejection treatment was initiated (for rejectors). Rejection refers to biopsy proven rejection, grade IA or greater as determined a staff pathologist PBMCs were purified from peripheral blood samples via density gradient centrifugation (cell preparation tubes, BD Pharmingen) and cryopreserved at −80 degrees C for future intracellular and extracellular staining and analysis via flow cytometry.

Ex vivo frequency and phenotypic analysis of isolated PBMCs

Standard extracellular staining was performed on PBMCs using the following fluorophore-labeled antibodies: CD14/CD19-V500 (BD Pharmingen), (Biolegend), CD8-BV786 (BD Pharmingen), CD28-PerCP-Cy5.5 (Biolegend), CCR7-PE-Cy7 (Biolegend), CD25-APC (Biolegend), CD45RA-qDot655 (Life Technologies), and CD4-APC-H7 (BD Pharmingen).

Ex vivo intracellular cytokine staining

For determination of ex vivo cytokine production, PBMCs were suspended in 1640 RPMI medium supplemented with 10% heat-inactivated fetal bovine serum (FBS), 1% L-glutamine (200mM), 1% penicillin/streptomycin (100×), 1% Hepes (1M), 1% 2-ME (14.3M). For T cell stimulation, 1 × 10⁶ PBMCs were placed in a 96 well plate and stimulated for 4 hours at 37 degrees C with a mixture of PMA (Sigma) and ionomycin (Sigma) at a concentration of 1 μg/mL each. Brefeldin A (GolgiPlug, BD Biosciences) was also added to all cells at a concentration of 1 μg/mL. Intracellular staining was performed after fixation and permeabilization according to manufacturer’s instructions (BD Biosciences) utilizing fluorophore-labeled antibodies to IFNγ-FITC and anti-IL-2 –AlexaFluor700 (Pharmingen) following extracellular staining as described above.

Statistics

Statistical analysis for flow cytometric assays done with 3 groups was performed using non-parametric one-way ANOVA with Prism 5.0 (GraphPad) software. Statistical analysis for flow cytometric assays with two groups was performed using non-parametric 2-tailed paired (Wilcoxon) or unpaired (Mann-Whitney) procedures as appropriate with Prism 5.0 (GraphPad) software. Tree analysis was performed with R software. For the clinical data continuous variables are summarized as median and range and compared using Mann-Whitney U test; categorical variables are expressed as relative and absolute frequencies, and compared using Chi-square test. Sensitivity and specificity were calculated using MedCalc. The statistical analysis was performed with SPSS version 21 for Windows (SPSS Inc., Chicago, IL). P values of less than 0.05 were considered statistically significant.

Results

Decreased frequencies of CD4⁺ CD28^null cells at baseline in patients who go on to experience acute rejection on belatacept-based immunosuppression

In order to determine immunophenotypic profiles of patients who go on to experience acute rejection following treatment with belatacept vs. those that remain stable, we enrolled renal transplant recipients receiving a belatacept-based immunosuppressive regimen at Emory University Hospital in an IRB-approved immune monitoring protocol. PBMC were isolated and banked prior to transplantation (baseline). We then performed a retrospective analysis of immune profiles by flow cytometry of patients who remained stable for the first year following transplant (n=13) compared to those that went on to experience an episode of acute rejection (n=10). As shown in Table 1, stable and rejector cohorts were not different with regard to age, gender, recipient CMV status, time on dialysis, or underlying renal disease.

Table 1.

Transplant Recipient Characteristics for Belatacept Treatment Regimen

	Stables (n=13)	Rejectors (n=10)	p value
Median age (years)	51 (39–72)	55 (25–79)	0.61
Male/female (%)	11 (84.6)/2 (15.4)	7 (70)/3 (30)	0.73
Cause of ESRD
Hypertensive nephrosclerosis	4	4	0.98
Diabetes	5	2	0.61
Glomerulonephritis	4	3	0.67
PCKD	0	1	0.89
CMV recipient status + (%)	11 (84.6)	7 (70)	0.73
Median time on dialysis (years)	5 (0.2–12)	5 (1.4–14)	0.98
Median time to rejection or follow up (months)	1 (1–5)	1.3 (0.5–7)	0.66

^*Continuous variables: median, range. Categorical variables: absolute number and frequencies. CMV: cytomegalovirus; ESRD: end stage renal disease; PCKD: polycystic kidney disease

To assess pre-transplant CD28 expression patterns within the CD4⁺ and CD8⁺ T cell compartments of patients that remained stable following renal transplantation and belatacept-based immunosuppression vs. those that did not, PBMC were analyzed by flow cytometry. Data shown are gated on lymphocytes, CD19/CD20⁺ cells were excluded, and cells were gated based on CD4 and CD8 expression (Figure 1A). CD28 expression on these cells was then determined. Surprisingly, pre-transplant samples of patients who went on to experience acute rejection while on belatacept exhibited decreased frequencies of CD28^null CD4⁺ T cells as compared to those that remained stable while on belatacept-based therapy. It is important to note that the frequency of CD28^null CD4⁺ T cells in stable patients was actually higher than that observed in normal healthy controls (Figure 1C), suggesting that the acquisition of increased frequencies of CD28^null CD4⁺ T cells, potentially as a consequence of chronic kidney disease, is associated with a more senescent, less allo-aggressive immune phenotype. In contrast, patients who went on to experience acute rejection instead possessed increased frequencies of CD28⁺ CD4⁺ T cells relative to those that remained stable (Figure 1C). The CD28 profiles of CD4⁺ T cells in these recipients looked more like that of normal healthy controls. Interestingly, these differences at the bulk T cell level were confined to the CD4⁺ T cell compartment, as no differences in the frequencies of CD28^null cells within the CD8⁺ T cell compartments of stables vs. rejectors were observed (Figure 1D–E).

Figure 1. Pre-transplant CD28 expression within the CD4⁺ and CD8⁺ T cell compartments of patients that remained stable on belatacept-based immunosuppression vs. those that went on to reject.

(A) Representative flow plots of the gating strategy to identify CD28^null CD4⁺ and CD8⁺ T cells. (B) Representative flow plot of pre-transplant CD28^null CD4⁺ expression in patients that remained stable while on belatacept-based therapy and those that rejected. (C) Patients who went on to experience acute rejection while on belatacept exhibited decreased frequencies of CD28^null CD4⁺ T cells as compared to those that remained stable while on belatacept-based therapy (p=0.03), and the frequency of CD28^null CD4⁺ T cells in stable patients was actually higher than that observed in normal healthy controls (p=0.01). Patients who went on to experience acute rejection possessed increased frequencies of CD28⁺ CD4⁺ T cells compared to those that remained stable (p=0.03) D. Representative flow plots of pre-transplant CD28^null CD8⁺ expression in patients that remained stable while on belatacept-based therapy and those that rejected. E. No differences in the frequencies of CD28^null cells and CD28⁺ cells within the CD8⁺ T cell compartments of stables vs. rejectors were observed.

Patients who go on to experience belatacept-resistant rejection possess increased frequencies of CD28⁺ CD4⁺ T_EM, CD28⁺ CD4⁺ T_EMRA, and CD28⁺ CD8⁺ T_EM at baseline

Based on these initial results we sought to determine whether the differences in %CD28⁺ cells within the pre-transplant CD4⁺ T cell compartment in stables vs. rejectors could be ascribed to one particular memory T cell subset. PBMC from the patient cohorts described above were stained with CD45RA and CCR7 to delineate naïve (CD45RA⁺ CCR7⁺), central memory (T_CM, CD45RA- CCR7⁺), effector memory (T_EM, CD45RA- CCR7-), and effector memory-RA (T_EMRA, CD45RA+ CCR7-) subsets. The overall pre-transplant frequencies of naïve, T_CM, T_EM, and T_EMRA cells within either the CD4⁺ or CD8⁺ T cell compartments were not different between patients who went on to experience acute rejection vs. those that did not (Supplemental Figure 1A, B). However, when assessing the frequencies of CD28⁺ cells within each CD4⁺ memory T cell subset, we found that the pre-transplant frequencies of CD28⁺ cells among CD4⁺ naïve cells and CD4⁺ T_CM were similar in patients that went on to reject vs. those that did not (Figure 2A, 2B). In contrast, the pre-transplant frequency of CD28⁺ cells among CD4⁺ T_EM was very significantly increased in patients who went on to reject vs. those that did not (p<0.0001). We also found a statistically significant increase in the pre-transplant frequency of CD28⁺ cells among CD4⁺ T_EMRA cells in these patients (p=0.01), although the delineation was not quite as distinct. Frequencies of CD28⁺ cells among naïve, T_CM, T_EM, and T_EMRA CD4⁺ T cell compartments in healthy controls are depicted in Supplemental Figure 2A and 2B.

Figure 2. Pre-transplant frequencies of CD28⁺ CD4⁺ T_EM, CD28⁺ CD4⁺ T_EMRA, and CD28⁺ CD8⁺ T_EM are increased in patients who experienced belatacept-resistant rejection.

(A) Representative flow plots and (B) summary data of frequencies of CD28⁺ cells within naïve (CCR7⁺ CD45RA⁺), T_CM (CCR7⁺ CD45RA-), T_EM (CCR7- CD45RA-), and T_EMRA (CCR7- CD45RA⁺) subsets of CD4⁺ T cells in patients that experienced belatacept-resistant rejection (n=10) vs. those that remained stable (n=13). Pre-transplant frequency of CD28⁺ cells among CD4⁺ T_EM and CD4⁺ T_EMRA cells were significantly increased in patients who went on to reject vs. those that did not (p<0.0001 and p=0.01, respectively). (C) Representative flow plots and (D) summary data of frequencies of CD28⁺ cells within naïve (CCR7⁺ CD45RA⁺), T_CM (CCR7⁺ CD45RA-), T_EM (CCR7- CD45RA-), and T_EMRA (CCR7- CD45RA⁺) subsets of CD8⁺ T cells in patients that experienced belatacept-resistant rejection (n=10) vs. those that remained stable (n=13). Pre-transplant frequency of CD28⁺ cells among CD8⁺ T_EM were significantly increased in patients who went on to reject vs. those that did not (p=0.05).

Because we failed to detect any difference in pre-transplant CD28⁺ T cell frequencies in stables vs. rejectors within the CD8⁺ T cell compartment, we next asked whether drilling down into the CD8⁺ memory T cell subsets would uncover any differences. As was done with the CD4⁺ compartment above, CD8⁺ T cells were gated into naïve, T_CM, T_EM, and T_EMRA subsets and the frequencies of CD28⁺ cells within each of those subsets was determined. These analyses revealed a statistically significant increase in the pre-transplant frequency of CD28⁺ cells within the CD8⁺ T_EM subset in patients that went on to reject vs. those that did not (p=0.05). Frequencies of CD28⁺ cells among naïve, T_CM, T_EM, and T_EMRA CD8⁺ T cell compartments in healthy controls are depicted in Supplemental Figure 2C and 2D.

Frequency of CD28⁺ CD4⁺ T_EM fails to stratify stables vs. rejectors among tacrolimus-treated renal transplant recipients

Given the above results, we next sought to determine whether this signature of higher frequencies of CD28⁺ CD4⁺ T_EM also functioned to stratify risk of rejection in patients treated with calcineurin inhibitor-based immunosuppression. As such, we assessed pre-transplant immune profiles in 15 patients who had previously undergone renal transplantation at Emory University Hospital and had been enrolled in an IRB-approved immune monitoring protocol (Table 2). In contrast to what we observed in patients who received belatacept-based immunosuppression, both pre-transplant %CD28⁺ of CD4⁺ and %CD28⁺ of CD4⁺ T_EM failed to stratify rejectors (n=7) vs. stables (n=8) in patients who received tacrolimus-based immunosuppression (Figure 3A, 3B). Likewise, pre-transplant assessment of the frequencies of CD28⁺ CD8⁺ T cells, either in the bulk or T_EM compartment, also failed to correlate with risk of rejection (Figure 3C, 3D). These findings suggest that the association of CD28⁺ CD4⁺ T_EM with increased risk of rejection is specific to costimulation blockade-based regimens.

Table 2.

Transplant Recipient Characteristics for Tacrolimus Treatment Regimen

	Stables (n=8)	Rejectors (n=7)	p value
Median age (years)	49.5 (31–60)	47 (43–63)	0.84
Cause of ESRD
Hypertensive nephrosclerosis	3	1	0.66
Diabetes	0	1	0.94
Glomerulonephritis	3	3	0.75
PCKD	2	2	0.66
CMV recipient status + (%)	5 (75)	5 (71.4)	0.85
Median time on dialysis (years)	2.5 (0–18)	4 (0–5)	0.54
Median time to rejection or follow up (months)	4 (0.9–5)	4.5 (0.9–6)	0.58

^*Continuous variables: median, range. Categorical variables: absolute number and frequencies. CMV: cytomegalovirus; ESRD: end stage renal disease; PCKD: polycystic kidney disease

Figure 3. Pre-transplant frequencies of CD28⁺ CD4⁺ or CD8⁺ T cells are not increased in patients who experienced rejection following treatment with tacrolimus-based immunosuppression vs. those that remained stable.

(A and B) Summary data of frequencies of CD28⁺ cells within the total CD4⁺ compartment (A) and T_EM subset (CCR7- CD45RA-) (B) of CD4⁺ T cells from PBMC of patients that experienced rejection during treatment with tacrolimus (n=7) vs. those that remained stable (n=8). (C and D) Summary data of frequencies of CD28⁺ cells within the total CD8⁺ compartment (A) and T_EM subset (CCR7- CD45RA-) (B) of CD8⁺ T cells from PBMC of patients that experienced rejection during treatment with tacrolimus (n=7) vs. those that remained stable (n=8). p=ns for all pairs.

Utility of frequency of CD28⁺ CD4⁺ T_EM as a biomarker to identify patients who are most likely to experience belatacept-resistant rejection

In order to investigate the statistical utility of this parameter as a potential biomarker for the pre-transplant stratification of patients likely to experience acute rejection while being treated with belatacept-based immunosuppression vs. those that are likely to remain stable, we conducted decision tree analysis (R) where variables including %CD28⁺ of CD4⁺ T cells, %CD28⁺ of CD4⁺ T_EM and %CD28⁺ of CD4⁺ T_EMRA were included in the analysis using this cohort of 22 patients undergoing renal transplantation followed by belatacept-based immunosuppression. The analysis selected the %CD28⁺ of CD4⁺ T_EM as the variable that could most effectively stratify stable versus rejectors. We determined that using a cutoff value of >83.6% CD28⁺ of CD4⁺ T_EM would allow us to identify 12/12 true stables and 8/10 true rejectors, generating a sensitivity of 80% (95% CI 44.39% to 97.48%) and a specificity of 100% (95% CI 73.54% to 100.00%).

CD28⁺ CD4⁺ T cells exhibit increased IL-2 production and rapidly convert to CD28⁻ phenotype following in vitro stimulation

We next sought to understand mechanisms underlying the association of increased CD28⁺ CD4⁺ T_EM cells with belatacept-resistant rejection. We interrogated the functionality of CD28⁺ CD4⁺ T_EM vs. CD28^null CD4⁺ T_EM subsets (both subsets were isolated from normal healthy controls). PBMC were stimulated ex vivo with PMA/ionomycin as described in the Materials and Methods, and CD4⁺ cells were gated on CD28⁺ vs. CD28⁻ and analyzed for IFN-γ and IL-2 production. As shown in Figure 4A, a high frequency of CD28^null CD4⁺ T cells secreted IFN-γ, but virtually no cells were capable of secreting IL-2, a critical T cell growth factor. In contrast, CD28⁺ CD4⁺ T cells were capable of secreting both IFN-γ and IL-2 (Figure 4A–C). Thus, these data demonstrate that CD28^null CD4⁺ T cells may be defective in their capacity to execute a successful rejection response driven by IL-2. Given these differences in the functional capacities of CD28^null vs. CD28⁺ CD4⁺ T cell subsets, we hypothesized that CD4⁺ populations isolated from stables vs. rejectors would exhibit differential cytokine production. To test this, pre-transplant PBMC from stables vs. rejectors were restimulated ex vivo and IFN-γ and IL-2 production was assessed. Stables and rejectors did not differ with regard to the frequencies of IFN-γ-secreting cells within their CD4⁺ T cell compartments (Figure 4D). However, patients who went on to reject exhibited a statistically significantly higher frequency of IL-2-secreting CD4⁺ T cells as compared to patients that experienced freedom from rejection (Figure 4E).

Figure 4. CD28⁺ CD4⁺ T_EM produce more IL-2 and rapidly downregulate CD28 following ex vivo restimulation.

(A–C) PBMC from healthy controls were stimulated ex vivo with PMA/ionomycin for 4 hours and intracellular cytokine staining was performed. CD4⁺ cells were gated on CD28⁺ vs. CD28⁻ and analyzed for IFN-γ and IL-2 production. Representative flow plots are shown in A. B–C, Significantly higher frequencies of CD28⁺ CD4⁺ T cells secreted IL-2 (but not IFN-γ) relative to CD28^null CD4⁺ T cells (n=3, p<0.05). (D) PBMC isolated from belatacept-treated stables and rejectors did not differ with regard to the frequencies of IFN-γ-secreting cells within their CD4⁺ T cell compartments following ex vivo restimulation. (E) Belatacept-treated patients who rejected exhibited a statistically significantly higher frequency of IL-2-secreting CD4⁺ T cells as compared to patients that were rejection-free (p<0.05). (F–G) PBMC from belatacept-treated rejectors vs. stables were stimulated ex vivo for 4 hours and the frequencies of CD28^null cells following ex vivo stimulation were quantified. F, The frequencies of CD28⁻CD4⁺ T cells were not different after ex vivo stimulation in patients that were stable. (G) The frequencies of CD28⁻CD4⁺ T cells significantly increased after ex vivo stimulation in patients that experienced belatacept-resistant rejection.

Data from our complementary NHP studies have also revealed an enrichment of CD28^null cells in rejecting allografts, suggesting that these may be the cells that actually mediate costimulation blockade-resistant rejection. We next queried the extent of CD28 downregulation following TCR stimulation in CD28⁺ CD4⁺ T_EM in stables vs. rejectors. Following ex vivo stimulation with PMA/ionomycin, CD4⁺ T cell populations from rejectors, which initially contained low frequencies of CD28^null cells, rapidly acquired a significant CD28^null population (Figure 4G). This rapid emergence of a CD28^null CD4⁺ T cell population was not observed in pre-transplant CD4⁺ T cell populations isolated from stable recipients (Figure 4F). These data demonstrate that CD28⁺ T cell populations are capable of rapidly generating a CD28^null component following TCR stimulation, as would be the case upon in vivo encounter with antigen.

CD28^null CD4⁺ T_EM isolated from stables but not rejectors expressed increased levels of the coinhibitory molecule 2B4

We next sought to identify phenotypic markers in CD28^null CD4⁺ T cell populations that might underlie the observed deficiency of these cells to optimally produce IL-2. Importantly, CD28^null CD4⁺ T cells were found to exhibit high expression of the T cell coinhibitory molecule 2B4 (Figure 5A). 2B4 is a member of the CD2 family and associates intracellularly with SHP-1 phosphatases to dampen T cell responses (28, 29), and we recently reported that upregulation of 2B4 was functionally important for the efficacy of CD28 costimulation blockade to inhibit alloreactive T cell responses in a murine model of transplantation (30). In this cohort of human renal transplant recipients, 2B4 expression was low (~5–10%) on CD28⁺ CD4⁺ T_EM, and was not different between stable patients and rejectors (Figure 5B). In contrast, stable patients exhibit statistically significantly higher expression of 2B4 on CD28^null CD4⁺ T cells compared to either rejectors or to healthy controls (Figure 5C). We observed no differences in the expression of coinhibitory molecules PD-1 and TIM-3 on CD28^null CD4⁺ T cells in stables compared to rejectors (data not shown). These data suggest that upregulation of 2B4 on CD28^null CD4⁺ T cells is associated with freedom from rejection following treatment with belatacept.

Figure 5. 2B4 coinhibitory molecule expression was increased on CD28^null cells in belatacept-treated stables relative to rejectors.

(A) Representative flow plots demonstrating that 2B4 is highly enriched on CD28^null cells relative to CD28⁺ CD4⁺ T cells (B) No statistical difference was observed in 2B4 expression on CD28⁺ CD4⁺ T_EM in PBMC isolated from belatacept-treated patients that were stable versus those that experienced rejection. (C) 2B4 expression was significantly higher in CD28^null CD4⁺ T_EM in PBMC isolated from belatacept-treated patients that were stable versus either those that experienced rejection or healthy controls (p=0.05).

ESLD patients posses increased frequencies of CD28^null CD4⁺ T cells similar to ESRD patients

Our data suggested that belatacept-treated patients that were free from rejection at one year exhibited significantly increased pre-transplant frequencies of CD28^null CD4⁺ T_EM relative to normal healthy controls. We thus queried whether end-stage renal disease (ESRD) is a driver of the CD28^null CD4⁺ T_EM phenotype. We identified a statistically significant increase in CD28^null cells in ESRD patients as compared to normal healthy controls within the CD4⁺ T cell compartment (Figure 6A). In contrast, CD28^null cells were not increased within the CD8⁺ T cell compartment of ESRD patients as compared to normal healthy controls (Figure 6B). In order to determine whether this increase in CD4⁺ CD28^null cells was common to chronic, end-stage organ failure diseases or was specific to ESRD, we interrogated the CD28 phenotypes in a cohort of end-stage liver disease (ESLD) patients awaiting liver transplantation (cohort characteristics described in Supplemental Table 1). Interestingly, ESLD patients also exhibited a significant increase in CD28^null CD4⁺ (Figure 6A) but not CD28^null CD8⁺ T cells (Figure 6B) relative to healthy controls, suggesting that the inflammatory milieu and immune dysregulation associated with ESLD also drives accumulation of CD28^null CD4⁺ T cells in some patients. 55% of this cohort of patients awaiting liver transplant exhibited frequencies of CD28⁺ CD4⁺ T_EM similar to those observed in the cohort of renal transplant recipients that experienced rejection on belatacept (Figure 6C).

Figure 6. Assessment of CD28 expression in patients with ESRD compared to patients with ESLD and healthy controls.

(A) PBMC isolated from healthy controls (n=8), patients with ESRD (n=37) and ESLD (n=9) were assessed for frequencies of CD28⁺ cells within both the CD4⁺ (A) and CD8⁺ (B) T cell compartments. Both ESRD and ESLD presented a statistically significant increase in CD28^null cells as compared to normal HC within the CD4⁺ T cell compartment (p=0.03 and p=0.04 respectively). (B) No difference in the frequencies of CD28^null cells in the CD8⁺ T cell compartment was observed between patients with ESRD, ESLD and HC. (C) PBMC isolated from patients with ESLD were gated on CD4⁺ T_EM (CCR7- CD45RA-) and frequencies of CD28^null cells were assessed.

Discussion

In this study we retrospectively assessed immune profiles of belatacept-treated renal transplant recipients that went on to experience acute rejection following transplantation vs. those that did not, and identified a population of CD28⁺ CD4⁺ T_EM that was significantly increased in pre-transplant PBMC samples isolated from rejectors vs. stables. These data suggest that pre-transplant frequency of CD28⁺ cells within the CD4⁺ T_EM subset could be used as a predictive biomarker to identify patients that are at increased risk of rejection with a belatacept-based immunosuppressive regimen. Because this test can be easily run by any clinical flow cytometry lab (as opposed to more onerous tests for donor-specific memory T cells that are currently not able to be performed at all transplant centers), it may be a useful addition to the transplant diagnostic armamentarium. Based on the data from this initial cohort, this biomarker would be predicted to have high sensitivity, in that 100% of patients that went on to experience rejection fell above the 83.6% cutoff for CD28⁺ cells as a percentage of CD4⁺ T_EM. However, this methodology would be predicted to classify 20% of rejectors as stables (i.e. a 20% false negative rate). Still, the potential promise of this biomarker to identify and eliminate 80% of patients who would go on to experience belatacept-resistant rejection warrants further testing in a larger, prospective cohort of renal transplant recipients.

While our data detected an association between high frequencies of CD28⁺ CD4⁺ T_EM and rejection, the study of course does not necessarily demonstrate a causal link between this population and the precipitation of rejection. Still, it is important to note that several mechanistic studies support a role for this cell type in mediating, or being associated with, belatacept-resistant rejection. For example, Espinosa et al. described a subset of CD57-expressing CD4⁺ T_EM that were associated with increased incidence of rejection while on belatacept (but not tacrolimus) treatment (31). While this study demonstrated that many CD57⁺ CD4⁺ cells exhibited a CD28^null phenotype, it also revealed that cytokine secretion and cytolytic function were much higher within the CD57⁺ CD4⁺ T cell population than the CD28^null CD4⁺ T cell population. Similarly, our data demonstrated that the CD28⁺ fraction of CD4⁺ T_EM cells were notable for their markedly higher secretion of IL-2, an important T cell growth factor that may be critical for the sustained response required to reject an allograft. Together, these findings suggest that perhaps a CD4⁺ CD28⁺ CD57⁺ T cell may be most capable of mediating (or is most highly associated with) belatacept-resistant rejection, a hypothesis that warrants further investigation. Another important study suggested that while CTLA-4 Ig inhibited proliferation of CD8⁺ CD28⁺ T cells in vitro, it had no effect on either CD8⁺ CD28^null or CD28⁺ memory T cell proliferation when IL-15 was added to the culture (18). These results implicate IL-15 as a potentially critical pathway that confers costimulation blockade resistance on CD28⁺ memory T cells. Because IL-15 is produced by renal epithelial during inflammation, it is interesting to speculate that blockade of IL-15-mediated signals may be sufficient to inhibit alloreactive memory CD28⁺ T cell responses and prevent belatacept-resistant rejection in human renal transplant recipients.

Our results also suggest that CD28⁺ CD4⁺ T_EM in rejectors are likely qualitatively different as compared to CD28⁺ CD4⁺ T_EM in stables, as these cells were better able to downregulate CD28 following in vitro activation, a finding which is consistent with more robust activation. Ongoing gene profiling experiments seek to understand qualitative differences between CD28⁺ CD4⁺ T_EM in belatacept-treated rejectors vs. stables. Another observation that may provide a clue regarding differential immunogenicity in stables is their high expression of the T cell coinhibitory molecule 2B4 on CD28^null CD4⁺ T_EM. Importantly, many studies of chronic viral infections in mice and humans have implicated 2B4 as a reliable marker of T cell exhaustion (28, 32–35), and a recent study that defined the cells that restore the immune response following PD-1 blockade identified them as being 2B4⁻ (36). Further, we recently implicated 2B4 as a critical pathway in the mechanism by which CD28 blockade attenuates alloreactive CD8⁺ T cell responses in mice (30). Thus, we postulate that possessing a relatively large compartment of impotent, 2B4-expressing T cells may result in a reduction in the precursor frequency of competent donor-reactive T cells below the threshold necessary to mediate costimulation blockade resistant-rejection. Indeed, we have previously shown that high precursor frequency is a critical determinant of susceptibility vs. resistance to a costimulation blockade based regimen (37). On the other hand, an alternative hypothesis is that 2B4⁺ CD28^null CD4⁺ T_EM are functioning to suppress the donor-specific immune response. Future studies are aimed at investigating these possibilities.

While our study identified a biomarker that is potentially predictive of rejection while on a belatacept-containing regimen, it is important to consider that patients in this cohort also received tacrolimus at early time points following transplantation, weaning at 3–9 months post-transplant (Adams et al., AJT 2016, In press). Many patients rejected during this wean period, leading us to conclude that these rejection episodes are precipitated by costimulation-independent T cell populations that become activated as CNI levels decline. Indeed, in our study the relative frequency of CD28⁺ CD4⁺ T_EM did not segregate rejectors vs. stables in tacrolimus-treated patients. Further, our conclusion that the frequency of CD28⁺ CD4⁺ T_EM is a biomarker specific to belatacept-resistant rejection is validated by a contemporaneous NHP study in which rhesus macaque renal transplant recipients that went on to experience rejection following treatment with a tacrolimus-free belatacept-containing regimen exhibited higher frequencies of CD28⁺ CD8⁺ memory T cells relative to non-rejecting controls (Mathews et al., submitted manuscript). Similar to our human study, this association of CD28⁺ memory T cells with rejection was not observed in the tacrolimus-treated control arm. Taken together, these data suggest that the mechanisms of rejection at play in recipients containing high pre-transplant frequencies of CD28⁺ CD4⁺ T_EM are likely effectively controlled by calcineurin inhibition. Identifying other immunologic pathways critical for recall responses in these CD28-independent populations remains an important step toward the goal of developing CNI-free treatment regimens for these patient populations.

As mentioned above, complementary data from a rhesus macaque model of renal transplantation also identified increased frequencies of CD28⁺ memory T cells as being predictive of belatacept-resistant rejection, however in this study the difference was localized to the CD8⁺ T_EMRA subset. While our study did identify a statistically significant increase in CD28⁺ T_EM in human rejectors vs. stable patients, the biggest differences were observed in the CD4⁺ T cell compartment. This difference between the NHP and human data may be related to the fact that the NHP recipients are immunologically more akin to healthy controls than uremic patients experiencing years of chronic kidney disease, which our data show results in an increase in the frequencies of CD4⁺ but not CD8⁺ CD28^null cells relative to human controls (Figure 6). These data highlight the complex nature of modeling the immunology of renal transplant recipients using animal systems, and suggest that recapitulating the uremic environment of chronic kidney disease in animal models may shed new light on alloimmunity and the mechanisms of allograft rejection.

Finally, it is interesting to note that similar to the increase in CD4⁺ CD28^null cells observed in ESRD patients awaiting kidney transplantation, a subset of ESLD patients awaiting liver transplantation also exhibited a significant increase in the frequency of CD28⁺ CD4⁺ T_EM cells relative to normal healthy controls. The mechanisms underlying belatacept-resistant rejection in liver transplantation is an important area of research, because while on the surface the relative ease of tolerization may make liver transplantation seem like an ideal context for use of belatacept, early clinical trials of belatacept in liver transplantation revealed markedly high rates of rejection (47% in the belatacept treated arm vs. 9% in control CNI treated arm) (38). Thus, while the mechanisms of belatacept-resistant rejection during liver transplantation remain an important area of investigation, it is interesting to speculate that the frequency of CD28⁺ CD4⁺ T_EM could also be used as a biomarker to predict risk of rejection of liver transplant recipients following treatment with belatacept treatment, potentially paving a path to clinical implementation of belatacept in a select subset of liver transplant recipients with the most “permissive” immune profiles. In sum, our data offer insight into the immunologic environments associated with belatacept-resistant rejection in human renal transplant recipients and demonstrate an example of how pre-transplant immune profiling may allow for personalized immunotherapy to improve outcomes following transplantation.

Supplementary Material

FigS1-2-TableS1

Figure S1: The overall pre-transplant frequencies of naïve, TCM, TEM, and TEMRA cells within both the CD4+ (A) and CD8+ (B) T cell compartments were not different between patients who went on to experience acute rejection vs. those that did not.

Figure S2: The frequencies of naïve, TCM, TEM, and TEMRA cells within both the CD4+ (A, B) and CD8+ (C, D) T cell compartments of healthy control subjects (n=5) are depicted.

Table S1: Liver transplant recipient characteristics for tacrolimus treatment regimen.

Abbreviations

CKD

chronic kidney disease

CNI

calcineurin inhibitor

EBV

Epstein–Barr virus

ESLD

end-stage liver disease

ESRD

end-stage renal disease

MMF

mycophenolate mofetil

PBMC

peripheral blood mononuclear cell

PTLD

posttransplant lymphoproliferative disease