This study was approved by Beth Israel Deaconess Medical Center Institutional Review Board. At our institution, 65 years older are the ages when kidney transplant candidates are considered as elderly candidates. In total, 59 subjects were recruited into five groups: (1) Healthy controls, (2) <65 years of age with CKD not on dialysis, (3) ≥65 years of age with CKD not on dialysis, (4) <65 years of age on dialysis, (5) ≥65 years of age on dialysis. Healthy controls were defined as individuals aged >18 years old with an estimated glomerular filtration rate (eGFR) >60 and no albuminuria. CKD patients were defined as adults listed for kidney transplant, but not yet requiring renal replacement therapy. Dialysis patients were defined as adults listed for kidney transplant and currently requiring renal replacement therapy. Patients with active infections, known systemic autoimmune conditions, those who have previously received immunosuppressant therapy, chemotherapy or a prior transplant were excluded. Patients were approached at their Transplant clinic visits and consented. Blood was collected from each subject. Healthy controls were approached at their Pre surgical visits for their elective surgery procedures. The IRP was defined as CMV seropositivity and CD4/CD8 ratio <1 (58) or being in the highest quintile of CD8+ senescent (CD28–/CD57+) T cells (14).

A 100 μL aliquot of fresh whole blood was stained using a master mix of fluorescently-conjugated antibodies targeting the following human markers: CD3 (BD Biosciences SP34.2), CD20 (BD Biosciences L27), CD4 (BD Biosciences L200), CD8 (BD Biosciences RPA-T8), CD45RO (Beckman Coulter UHCL1), CD45RA (BD Biosciences 5H9), CD197/CCR7 (BD Biosciences 150503), CD28 (ThermoFisher 28.2), and CD57 (BD Biosciences NK-1). Following a 20 min/4°C staining incubation, automated red blood cell lysis was performed on the stained samples using a TQ-Prep™ Workstation (Beckman Coulter). Samples were then twice washed with 1X Dulbecco's Phosphate-Buffered Saline (Gibco), and subsequently fixed for cytometric analysis in 1.5% formaldehyde. Multicolor flow cytometry was then performed on the samples using a LSRII flow cytometer (BD Biosciences), and results were analyzed using FlowJo software (Treestar).

Specific cell populations were gated according to the following strategy: T-cells were identified using forward vs. side scatter, excluded of doublets, and further gated on CD3+ cells using a CD20 vs. CD3 plot to exclude B-cells. To gate CD3+/CD4+ and CD3+/CD8+ subpopulations most accurately, a Boolean gating strategy was designed directly off the CD3+ parent population. The effector cell subpopulation was defined for the CD3+/CD4+ and CD3+/CD8+ populations using Boolean gating for the CD3+/CD197+/CD45RA-/CD45RO+ cells. In a similar manner, the effector memory (EM), central memory (CM), and naive subpopulations of the CD3+/CD4+ and CD3+/CD8+ populations were defined using Boolean gating for the CD3+/197CD–/CD45RA–/CD45RO+, CD3+/CD197+/CD45RA–/CD45RO+, and CD3+/CD197+/CD45RA+/CD45RO– cells, respectively. Finally, senescent cells were defined for each group by applying Boolean gating for CD3+/28CD–/CD57+ cells to all subpopulations defined above.

To analyze NK cells, a 100 μL aliquot of fresh whole blood was stained using a master mix of fluorescently-conjugated antibodies. NK cells were defined as lymphocytes that were 3CD– (BioLegend UCHT1), 14CD– (BioLegend M5E2), and 19CD– (BioLegend HIB19). NK cell subpopulations were then identified based on their expression of CD56 (BioLegend HCD56) and/or CD16 (Pharmingen 3G8), namely bulk NK cells including CD56^dim (3CD–CD56+CD16+), CD56^bright (3CD–CD56+16CD–), and CD56neg (3CD−56CD–CD16+). Additional antibodies targeting the following human markers were used to assess their expression in NK cells: CD25 (eBioscience BC96), NKG2C (R&D Systems 134591), KLRG1 (BioLegend 2F1), CD2+ (Pharmingen RPA2.10), CD57 (BioLegend HCD57), CD27 (BioLegend 0323), FcεRIγ- (Millipore Sigma). Samples were fixed and analyzed using a LSRII flow cytometer (BD Biosciences), and data analyzed using FlowJo software 7.6.5 (Treestar).

To quantify NK cell function, thawed peripheral blood mononuclear cells (PBMC)s were incubated at 1 × 10⁶ cells/mL in RPMI-1640 supplemented with 10% fetal bovine serum, 2 mM L-glutamine, 100 mg/mL streptomycin, 100 U/mL penicillin, and 1 ng/mL rhIL-15 without stimulus or in the presence of 2 μg/mL CMV pp65 peptide pool (NIH AIDS reagents), 2 μg/mL BK virus capsid protein VP1 peptide pool (15 mer overlapping by 11, JPT), a combination of 50 ng/mL rhIL-12+100 ng/mL rhIL-18, HLA-I-deficient K562 or 721.221 cells at effector-to-target ratio of 10:1. To measure NK-cell mediated antibody-dependent cellular cytotoxicity (ADCC) function, p815 cells (a mouse leukemic cell line) were co-cultured with 1 mg/ml p815-specific antibody (Abcam) for 1 h and then washed twice. PBMCs were then stimulated with p815 cells or p815 cells plus p815-specific antibodies at an effector-to-target ratio of 10:1. PBMCs were stimulated for 6 h in the presence of 5 uL/mL CD107a-BV786 antibody (H4A3, BD Biosciences) and 1 μL/mL GolgiStop (monensin; BD Biosciences) prior to surface staining using the following anti-human monoclonal antibodies from BD Biosciences: CD3 (SP34-2), CD14 (M5E2), CD19 (HIB19), CD56 (B159), CD16 (3G8), fixed, permeabilized (BD Biosciences), and finally stained for intracellular interferon using IFN-γ-FITC (B27; BD Biosciences). Multiparameter flow cytometric analysis was performed on an LSRII instrument (BD Biosciences).

The ImmuKnow® assay was performed by Viracor-Eurofin laboratory (Lee's Summit, MO) in accordance with published protocol (59). This assay determines total ATP released from CD4+ T cells with ex vivo mitogen stimulation by phytohemaglutinin-L. The assay has been approved by the United States Food and Drug Administration as a measurement total CD4+ T cell response in transplant recipients.

Data analysis was performed using STATA statistical software, version 14 (College Station, Texas). Continuous data was analyzed first utilizing the 5 original patient cohorts as predictors utilizing a Kruskal-Wallis test. Association between categorical variables was measured via Fisher's Exact test. In order to evaluate the contributions of each individual parameter on the outcome variables, univariable regression screen was performed. Any significant variables were placed into a multivariable linear regression. A partial least square discriminant analysis (PLSDA) using the mixomics R package (http://mixomics.org/methods/pls-da/) was performed to determine which features contribute to IRP. PLS-DA uses covariance to identify linear combinations of independent or latent variables that best differentiate between the different groups. Each variable is assigned a score, which can be visualized in the latent variable space (score plots). Latent variable loadings (loadings plots) can then be used to identify biomarker profiles associated with different groups. The prediction power of each set of variables was assessed using area under the curve (AUC) implemented in the mixomics package (60).

Patient clinical characteristics are compared in Table 1. There were no significant differences between the study groups except for their age and the group with <65 and on dialysis had a large proportion of patients with diabetes as the cause of their CKD. We analyzed the CKD stages for the two groups not on dialysis. Within each group about half was in stage IV and half in stage V. There was no significant difference between the two groups. As expected the estimated glomeruli filtration rate (eGFR) were significantly lower in the two groups on dialysis.

Based on the definition of IRP, as CMV seropositivity and CD4/CD8 ratio <1 (58) or being in the highest quintile of CD8+ senescent (28CD–/CD57+) T cells (14), weidentified 14 individuals as having the IRP (IRP+). The 14 individuals were equally divided amongst the 5 original study groups (P = 1.0). Of the 40 CMV seropositive patients, 14 were found to be IRP+ (35%). Of the patients in the highest quintile of CD8+ senescent cells, 0 were found to be CMV negative, which is significantly different from the other quintiles (P = 0.003). Caucasians were less likely to be IRP+ than black/AA (0.36; 0.15–0.58) or Asian (0.42; 0.09–0.75). No other factors including CKD/dialysis status, dialysis modality or age were associated. Diabetes was included due to its near significance on the univariable logistic regression. It should be noted that whites were significantly less likely to be CMV positive with only 17/39 (44%) being CMV positive compared to 16/19 (84%) blacks/AA and 5/6 (83.3%) Asians (P = 0.003) consistent with previous population descriptions (62). Racial disparities in IRP status persisted when comparing only CMV+ Caucasians to black/AA (0.35; 0.02–0.68), but did not persist for Asian patients. In order to test if any of the other patients features except those that were used to determine IRP (CMV status, CD4 and CD8, and CD8+ senescent (28CD–/CD57+) cells, could predict the IRP status, we performed a PLS-DA analysis using the IRP status as the outcome and all factors except those that were used to determine IRP, as predictors. This model was able to predict IRP with an accuracy of 1 and a p-value = 0.0004 (Figure 1) suggesting in addition to CMV status, CD4 and CD8, and CD8+ senescent (28CD–/CD57+) cells, other factors contribute significantly in defining IRP.

On multivariable analysis, age and IRP+ but not CMV+/IRP– status were found to affect frequency of CD4+. The percent of CD4+ T cells with a naïve phenotype decreased in both IRP+ and dialysis patients (Table 2A). Age did not affect the proportion of CD4+ naïve cells. The proportions of CD4+ CM cells remained unaffected by any of the variables. IRP+ status is strongly associated with an increased proportion of CD4+ EM cells. The proportion of CD4+ effector cells were not influenced by CMV status, but IRP is associated with an increase of these cells and CKD associated with a decrease. The proportion of CD4+ T cells which are 28CD–/CD57+ indicating senescence did not differ between the 5 cohorts (Table 2B). They were substantially and only increased in IRP+ patients and were not affected by other factors. CD4+ naïve and CM senescent cells are very rare and seen essentially only in IRP+ patients. CD4+ EM and effector senescent cells are more common and significantly increased in IRP+ and CMV+/IRP– patients, but are not affected by age, CKD or dialysis status.

On multivariable analysis, IRP+ but not CMV+/IRP– was associated with an increase in CD8+ T-cells while age was associated with a decline (Table 3A). CD8+ naïve cells were reduced in both the CMV+/IRP– and IRP+ groups and were unaffected by age, CKD and dialysis status. CD8+ CM cells were decreased in IRP+ patients. CD8+ EM were not affected by any variable. CD8+ TEMRA cells were expanded in both CMV+/IRP– and IRP+ groups. The proportion of CD8+ T cells which are 28CD–/57+ indicating senescence was increased only in the IRP+ group, and slightly decreased with age (Table 3B). The CMV–/IRP+ group was not different from the CMV– group. CD8+ senescent cells were unaffected by CKD or dialysis status. CD8+ naïve senescent cells were rare and only decreased in IRP+ patients. CD8+ CM naïve senescent cells are also rare and slightly increased with age. CD8+ EM and TEMRA cells were markedly increased in IRP+ patients and not in CMV+/IRP– patients and were not affected by the other variables.

We next evaluated if enhanced proportions of NK cells with phenotypic features reflecting terminal differentiation are associated with CMV serostatus, age, kidney disease, and dialysis in our cohort of renal transplant candidates. To do so, we determined frequencies of NK cells expressing markers of NK cell activation (CD25), as well as those indicative of differentiation (CD57, KLRG1, NKG2C, CD2, and CD27) or adaptive features (FcεRIγ) (Tables 4A,B). Multivariable analysis demonstrates that age is associated with an increase in NK cells, which is not affected by IRP or CKD status. There were no differences in CD2, CD25, or CD57 expression. CMV+/IRP– and IRP+ status both tended toward an increase in CD27+ CD56^bright NK cells, though this did not reach significance. Memory-like FcγRΔg NK cells were significantly increased in CKD but not in dialysis patients, and were not significantly affected by IRP status or age. KLRG1+ NK cells were significantly reduced in IRP+ patients but were unaffected by the other factors. NKG2C+ NK cells were significantly increased in dialysis patients but not affected by the other factors. To examine if race affected NK cell phenotypes in our cohort, we performed analysis of data from the Caucasian subjects only and did not show additional significant difference.

To further determine if NK cell profiles can discriminate IRP+ and IRP– patients and can be used in combination with ImmuKnow® values as a surrogate predictor of IRP, we performed a PLSDA analysis using NK cell phenotypic and functional data alone and in combination with ImmuKnow® values. Our multivariate analysis revealed that NK cell features, including proportions of bulk NK cells, and percentages of NK cells expressing KLRG1, CD27, CD57, and NKG2C, predicted IRP with an accuracy of 0.78 and a p-value of 0.002 (Figure 2A). When using ImmuKnow® alone we achieved an accuracy of 0.68 and p-value of 0.04 (Figure 2B). Interestingly, by combining the ImmuKnow® variable and NK cell features, we enhanced the prediction power of our model by reaching an accuracy of 0.83 and a p-value of 0.0002 (Figure 2C). This analysis showed that combining NK features and ImmuKnow® segregated IRP+ and IRP– groups where 18% of the variability between these two groups was explained by the first component and 11% by the second component of the PLSDA (Figure 2D). Several NK features were significant on both components and associated with either IRP+ or IRP– (Figures 2E,F).

Overall, this analysis showed that NK cell signatures contributed significantly to the IRP status where KLRG1, FcγRΔg NK cells, and CD27 were associated with the IRP– status and CD2, CD57, NKG2C, and proportions of NK cells were associated with the IRP+ status (Figure 2G). Furthermore, NK cell function against BK or CMV peptide pools also contributed to the NK cell-driven differentiation of IRP, with the IRP+ status being associated with IFN-γ production and the IRP– status with degranulation, as measured by CD107a upregulation, in response to those viral antigens (see Supplementary Material).

The ImmuKnow® assay was not significantly different across the 5 study groups (P = 0.08). In order to evaluate the role of CMV positivity as well as the IRP, we performed further analysis comparing CMV+ patients who did not display the IRP (CMV+/IRP–) and IRP+ to CMV– patients (Table 5). A multivariable analysis demonstrated that increased age (−4.70; P = 0.005) was significantly associated with decreased ImmuKnow® value (Table 5). Dialysis patients tended to have a higher ImmuKnow® value (111.40; p = 0.08), and CMV seropositivity was associated with lower ImmuKnow® value (−74.21, p = 0.16), though these did not reach significance. Other demographic and clinical factors, including CKD not requiring dialysis, were not associated with ImmuKnow® values.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.