Terminally Differentiated Effector Memory CD8 T Cells Identify Kidney Transplant Recipients at High Risk of Graft Failure

CLINICAL RESEARCH www.jasn.org

1 Terminally Differentiated Effector Memory CD8 T Cells Identify Kidney Transplant Recipients at High Risk of Graft Failure

Lola Jacquemont,1,2 Gaëlle Tilly,1,2 Michelle Yap,1,2 Tra-My Doan-Ngoc,1,2 Richard Danger ,1,2 Pierrick Guérif,2 Florent Delbos,3 Bernard Martinet,1,2 Magali Giral,1,2 Yohann Foucher ,4 Sophie Brouard,1,2 and Nicolas Degauque 1,2

Due to the number of contributing authors, the afﬁliations are listed at the end of this article.

ABSTRACT Background Identifying biomarkers to predict kidney transplant failure and to define new therapeutic targets requires more comprehensive understanding of the immune response to chronic allogeneic stimulation. 1 Methods We investigated the frequency and function of CD8 Tcellsubsets—including effector memory 1 (EM) and terminally differentiated EM (TEMRA) CD8 Tcells—in blood samples from 284 kidney transplant 1 recipients recruited 1 year post-transplant and followed for a median of 8.3 years. We also analyzed CD8 T cell reactivity to donor-specific PBMCs in 24 patients who had received living-donor kidney transplants. 1 Results Increased frequency of circulating TEMRA CD8 T cells at 1 year post-transplant associated with increased risk of graft failure during follow-up. This association remained after adjustment for a previously reported composite of eight clinical variables, the Kidney Transplant Failure Score. In contrast, increased 1 1 frequency of EM CD8 T cells associated with reduced risk of graft failure. A distinct TEMRA CD8 T cell subpopulation was identified that was characterized by expression of FcgRIIIA (CD16) and by high levels of proinflammatory cytokine secretion and cytotoxic activity. Although donor-specific stimulation induced a 1 similar rapid, early response in EM and TEMRA CD8 T cells, CD16 engagement resulted in selective 1 activation of TEMRA CD8 T cells, which mediated antibody-dependent cytotoxicity. 1 Conclusions At 1 year post-transplant, the composition of memory CD8 T cell subsets in blood improved prediction of 8-year kidney transplant failure compared with a clinical-variables score alone. A subpopu- 1 lation of TEMRA CD8 T cells displays a novel dual mechanism of activation mediated by engagement of 1 the T-cell receptor or of CD16. These findings suggest that TEMRA CD8 T cells play a pivotal role in 1 humoral and cellular rejection and reveal the potential value of memory CD8 T cell monitoring for predicting risk of kidney transplant failure.

JASN 31: 876–891, 2020. doi: https://doi.org/10.1681/ASN.2019080847

Because kidney transplantation remains the best Received August 27, 2019. Accepted January 16, 2020. therapeutic option for ESKD, and considering L.J. and G.T. contributed equally to this work. that current immunosuppression approaches inef- ﬁciently treat chronic rejection, favorable kidney Published online ahead of print. Publication date available at www.jasn.org. graft survival requires the design of innovative pre- ventive tools and therapeutics adapted to patients’ Correspondence: Dr. Nicolas Degauque, Centre de Re- cherche en Transplantation et Immunologie, UMR 1064, individual risks. A better understanding of the im- Hôtel Dieu–CHU de Nantes, 30 Boulevard Jean Monnet, mune response resulting from chronic allogeneic Loire-Atlantique, 44093 Nantes Cedex 01, France. Email: stimulation is thus needed (1) to identify novel bio- [email protected] markers that anticipate the risk of allograft injury, Copyright © 2020 by the American Society of Nephrology

876 ISSN : 1046-6673/3104-876 JASN 31: 876–891, 2020 www.jasn.org CLINICAL RESEARCH and (2) to provide new perspectives on the identification of Significance Statement new therapeutic targets to prolong allograft survival. The characterization of the humoral response has been Identifying biomarkers for predicting kidney transplant failure requires heavily scrutinized over the last decades and has led to inten- better understanding of the immune response to chronic allogeneic stimulation. The authors demonstrated that 1 year after kidney sive development of monitoring tools that detect the presence 1 fi transplantation, the composition of CD8 memory T cell subsets in of donor-speci c anti-HLA antibodies and C1q binding activ- blood—specifically the ratio of terminally differentiated effector – 1 ity as markers of kidney graft rejection.1 3 Quantification of memory (TEMRA) and effector memory CD8 T cells—is associated anti-HLA donor-specific antibody (DSA) using single-antigen with risk for subsequent graft failure and adds predictive value to a beads based on the mean fluorescence intensity clarified previously reported eight-variable clinical risk score. They also found 1 – that the intrinsic properties of DSAs (e.g., C1q-, C3d-, and that TEMRA CD8 T cells display a novel T cell receptor independent mechanism of activation that is mediated through CD16 engagement C4d-binding DSA IgG3 subclasses) are associated with an in- and results in inflammation and antibody-dependent cellular cyto- 1 crease in kidney graft failure beyond the titer of DSA.1,4,5 For toxicity. These findings suggest a pivotal role for TEMRA CD8 Tcells instance, Bouquegneau et al.6 reported an increase in allograft in chronic humoral and cellular rejection leading to kidney transplant 1 loss (hazard ratio [HR], 3.26) in patients with circulating failure. Future clinical benefits may include the use of CD8 memory T cell monitoring to improve risk prediction for graft failure and de- complement-activating DSA compared with that in patients 1 velopment of therapeutic strategies targeting TEMRA CD8 T cells. without complement-activating DSA in a large meta-analysis of 37 studies. In addition to complement-dependent mechanisms, the Fc segment of IgG may stimulate CD16-dependent and high levels of IFN-g, GZMB, and perforin (PERF-1).16–18 cytotoxicity and inflammatory functions such as those of in- Interestingly, similar attrition of TCRVb repertoire diversity was nate immune natural killer (NK) cells. An NK-cell molecular observed in KTx despite stable graft function for several years18 signature has been identified in biopsies from recipients of a and was shown to be associated with an expansion of TEMRA 1 2 kidney transplant (KTx) with antibody-mediated chronic re- (CD45RA CCR7 ) CD8 T cells.18,19 Studies with clinical jection (ABMR).7–9 Histologic analysis of biopsies from KTx follow-up showed that KTx with a high frequency of TEMRA 1 2 with C4d and C4d ABMR showed an enrichment of NK cell CD8 T cells exhibit a twofold higher risk of kidney dysfunction and macrophage infiltration according to CD56 and CD68 than those with a low frequency of TEMRA CD8 cells.19 How- expression compared with that in KTx with T cell–mediated ever, the factors that regulate the expansion and function of 1 rejection.7 Interestingly, similar infiltration by CD3 cells was TEMRA T cells, as well as their restriction toward donor anti- found in biopsies from KTx diagnosed with ABMR or gens, remain poorly defined. We recently provided evidence that Tcell–mediated rejection.7 Moreover, the NK-related gene IL-15 is a potent activator of TEMRA CD8 cells from KTx and signature differed between different studies,8–10 with an ab- healthy volunteers (HV)20 and that, upon IL-15 stimulation, sence of CD16 in some studies7–9; interestingly, the genes used TEMRA CD8 cells from KTx promote inflammation by induc- to define the NK signature (FGFBP2, SH21B, MYBL1, ing the expression of inflammatory CX3CL1/fractalkine by en- CX3CR1, KLRF1, GNLY,andCD16) were also overexpressed dothelial cells in an IFN-g– and TNF-a–dependent manner.20 by effector memory (EM) T cells expressing CD45RA (termi- Although risk factors of graft failure are well known, a nally differentiated EM [TEMRA]) CD8 T cells (Supplemental challenging issue in kidney transplantation is to predict out- Figure 1). This shared ABMR signature between NK and comes to help physicians guide patient care. In 2010, we cre- TEMRA CD8 cells prompted us to revisit the involvement of ated the Kidney Transplant Failure Score (KTFS),21 which is a CD8 T cells in the process of kidney allograft rejection. composite score calculated 1-year post-transplantation using Although complement-dependent mechanisms were first eight accepted clinical pre- and post-transplantation variables considered the main pathway leading to the pathogenic effects (recipient gender and age, donor creatinemia, number of pre- of alloantibodies, the role of the cellular immune response is vious transplantations, recipient creatinemia at 3 and 12 clearly not limited to the early post-transplant period but also months, recipient proteinuria at 12 months, and occurrence participates in the late humoral and cytotoxic responses of acute rejection within the first year). The KTFS is associated of chronic rejection. Memory T cells are considered a main with an area under the time-dependent receiver operating obstacle to achieving transplantation success and even trans- characteristic (ROC) curve (AUC) of 0.78 for a prognostic plantation tolerance. Preexisting memory T cells are associ- of kidney graft survival up to 8 years post-transplantation, ated with high incidence rates of severe rejection episodes,11 which is better than 1 year creatinemia or 1 year eGFR. The and KTx prone to acute rejection have a higher precursor KTFS is presented as a methodologic model of predictive frequency of alloreactive CD8 T cells than patients who were tools22 and is currently under clinical evaluation for its effi- nonrejectors.12 The involvement of CD8 T cells is not re- ciency to drive personalized follow-up by video conferencing stricted to early post-transplant events. KTx with biopsy- at home instead of face-to-face outpatient visits.23 In accor- proven ABMR exhibit high levels of the cytotoxic molecule dance with Braun et al.24 and Moore et al.,25 we believe that granzyme B (GZMB)13–15 within their grafts and an increase associating biologic—especially immunologic—biomarkers 1 2 in circulating CD8 CD28 T cells, with attrition of T cell re- to such complex clinical predictors could enhance their clin- ceptor (TCR) variable b-chain (TCRVb) repertoire diversity ical utility and improve personalization of patient follow-up.

JASN 31: 876–891, 2020 TEMRA CD8+ TCells 877 CLINICAL RESEARCH www.jasn.org

2 2 1 2 In this study based on an independent cohort of 284 KTx, we (CD45RA CCR7 ), and TEMRA (CD45RA CCR7 ) 1 1 showed that the measurement of CD8 memory subsets at 1 year CD3 CD8 cellsfromKTxwereFACSsorted(purity post-transplant, in association with the clinical metrics summa- .95%, FACSAria; BD Biosciences). Donor PBMCs were rized within the KTFS,21 allows improved identification of pa- T-cell depleted using a Pan T Cell Isolation Kit (Miltenyi) tients who will later return to dialysis as a consequence of graft and irradiated at 35 Gy. T cell–depleted donor PBMCs were failure. We demonstrated that donor-specific stimulation simi- cocultured with CPDeFluor405 KTx cells at a 1:2 ratio in 96-well, larly activates a rapid and early response of TEMRA and EM round-bottom plates at 37°C in 5% carbon dioxide (CO2)in 2 2 (CD45RA CCR7 ) CD8 T cells. Finally, we identified a unique the presence of anti-CD107a mAb (except for proliferation innate-like signature of TEMRA CD8 T cells using RNA se- assays). Cells were harvested and stained with anti-CD25 quencing (RNA-seq) and mass cytometry (CyTOF) approaches and anti-CD69 mAbs (day 1 and day 2) or with anti-CD3 and revealed a dual-trigger mechanism for TEMRA CD8 acti- mAb and Annexin V-FITC for proliferation assays (day 5). vation mediated by the engagement of TCR or CD16 pathways The production of 13 analytes was measured in culture super- that results in the secretion of proinflammatory cytokines and natant from CD8 T cell subsets using a LEGENDplex Human the activation of the associated cytotoxic response. CD8/NK Panel 13-plex (BioLegend).

CD16 Stimulation METHODS PBMCs were thawed and rested O/N in TexMacs medium, and CD8 T cells were purified using a CD8 T Cell Isolation Blood Samples Kit. Purified CD8 T cells were cultured for 6 hours in 96-well, fl PBMCs were separated from blood samples collected in EDTA at-bottomed plates at 37°C in 5% CO2 and, when indicated, tubes on a Ficoll gradient layer according to the manufac- treated with plate-bound anti-CD16 mAb (aCD16; 10 mg/ml; turer’s recommendations and were frozen in DMSO supple- clone 3G8; produced in house) and/or IL-15 (10 ng/ml; mented with 10% autologous serum. Miltenyi). Anti-CD107a phycoerythrin was added at culture initiation. Brefeldin A (10 mg/ml; Sigma) was added for Polychromatic Flow Cytometry 4 hours. After 6 hours of stimulation, cells were stained for The antibodies used for cytometric analyses are listed in cell surface markers (CD3, CD8, CD45RA, and CCR7) and, Supplemental Table 1. A total of 23106 frozen PBMCs were when indicated, for intracellular cytokines (TNF-a and IFN-g). surface stained with antibodies specific for CD3, CD8, CD45RA, CCR7, CD27, CD28, and CD57. In addition to Antibody-Dependent Cellular Cytotoxicity Response this cocktail, intracellular staining was performed using anti- The antibody-dependent cellular cytotoxicity (ADCC) re- bodies directed against T-bet, GZMB, PERF-1, and EOMES sponse of CD8 subsets was measured after challenge and cul- after fixation and permeabilization (Intracellular Fixation & ture with Raji cells coated with anti-CD20 antibody. CD8 Permeabilization Buffer Set; Thermo Fisher). A Yellow T cells were cultured for 30 minutes with Raji cells precoated LIVE/DEAD Fixable Dead Cell Stain Kit was used to exclude or not precoated with 10 mg/ml rituximab (Roche) at a ratio of dead cells from the analysis. BD CompBeads stained separately 10:1 in 96-well, round-bottomed plates at 37°C in 5% CO2 in with individual mAbs were used to define the compensation the presence of anti-CD107a phycoerythrin mAb. After 6 or matrix. Cells were analyzed with an LSRII flow cytometer 18 hours of stimulation, cells were stained for cell surface (BD Immunocytometry Systems). Data were analyzed using markers (CD3, CD8, CD45RA, and CCR7). FlowJo version 9.7.6 (TreeStar). For the assessment of CD16 expression by CD8 T cell subsets, PBMCs were stained Binding of CD8 to Single-Antigen HLA Class II Beads with antibodies against CD3, CD8, CD45RA, CCR7, and Purified CD8 T cells or cells from CD8 subsets 5 5 CD16. CyTOF data from Bengsch et al.26 were analyzed to in- (2310 –3310 ) were incubated O/N with 2 mlofsingle- vestigate the phenotype of CD16-expressing CD8 T lymphocytes. antigen HLA class II (OneLambda) beads previously incu- CD8 T lymphocytes were identified as iridium intercala- bated with 20 ml of serum from immunized KTx (treated 1 2 1 2 tor–positive, singlet LD–negative CD45 TCRgd CD3 CD4 with 0.1 mol/L EDTA for 10 min). When appropriate, cells 1 CD8a T cells and were then clustered according to the expres- were stained with anti-CD8, anti-CD45RA, and anti-CCR7 sion of 24 markers using Cytofkit and PhenoGraph.27–29 We then mAbs to identify CD8 subsets. A Boolean gating strategy was curated 23 high-dimensional clusters identified by PhenoGraph. used to assess the frequency of cells in CD8 subsets bound to single-antigen HLA class II beads in the presence of the FACS Sorting, Cell Culture, and Multiplex Cytokine appropriate serum (Supplemental Figure 2). Production Measurement RNA-Seq Mixed Lymphocyte Reaction 1 1 2 2 PBMCs were thawed, rested overnight (O/N) in TexMacs Me- Naive (CD45RA CD28 ), EM (CD45RA CD28 ), and 1 2 1 1 dium (Miltenyi), and stained for cell surface markers (CD3, TEMRA (CD45RA CD28 )CD3 CD8 T cells were FACS 1 1 CD8, CD45RA, and CCR7). Naive (CD45RA CCR7 ), EM sorted (purity .95%) from freshly isolated human PBMCs

878 JASN JASN 31: 876–891, 2020 www.jasn.org CLINICAL RESEARCH obtained from eight healthy donors (obtained by the Etablisse- used to correct for the multiplicity of the tests. The CD8 ment Francaiş du Sang). Cell pellets were resuspended phenotypes with a corrected P value ,0.05 were selected for in Buffer RLT (Qiagen) containing 1% b-mercaptoethanol further analyses. The discriminatory capacities were before subsequent RNA extraction using an RNeasy Micro evaluated by the AUC for data up to 8 or 11 years post- Kit according to the manufacturer’s instructions (Qiagen). transplant obtained via the inverse probability censoring The quality and quantity of the RNA were assessed by infrared weighted estimator.39 The corresponding 95% confidence in- spectrometry (Nanodrop) and an Agilent bioanalyzer (Agilent tervals (CIs) and P values related to the differences between RNA 6000 Pico Kit). Smart-Seq2 libraries were prepared AUC values were obtained by nonparametric bootstrap sam- by the Broad Technology Labs and sequenced by the Broad pling (1000 iterations). Genomics Platform according to the SmartSeq2 protocol with All statistical analyses were performed using R version 3.3.2 some modifications.30 Briefly, total RNA was purified using or GraphPad Prism. The package ROCt version 0.9 was used to RNA-SPRI beads, polyA1 mRNA was reverse transcribed to generate the time-dependent ROC curves (www.labcom-risca. cDNA, and amplified cDNA was subject to transposon-based com/packages-r). The package nricens was used to calculate fragmentation that used dual indexing to barcode each frag- the net reclassification improvement. The package corrplot ment of each converted transcript with a combination of (https://github.com/taiyun/corrplot) was used to calculate sample-specific barcodes. Sequencing was carried out as and visualize the correlation between the CD8 cell–related paired-end 2325-bp reads with an additional eight cycles populations. Mann–Whitney U tests, Kruskal–Wallis tests fol- per index. Data were separated by barcode and aligned using lowed by Dunn post hoc tests, and paired Wilcoxon tests were TopHat version 2.0.10 with the default settings. Transcripts used as appropriate, and the type of test used is included in were quantified by the Broad Technology Labs computational the figure legends. Multiple comparisons were corrected using pipeline using Cuffquant version 2.2.1.31 Briefly, data were the two-stage linear step-up procedure of Benjamini et al.,40 processed through Cuffnorm if 50% of the reads aligned and and Q was set to 5%. All P values are given as exact values or if at least 100,000 pairs were aligned per sample. The default as P,0.001. settings, including geometric normalization, were used for normalization and expression level data in the form of Study Approval log2-transformed fragments per kilobase of transcript per PBMCs were prospectively collected from 284 KTx in the million mapped fragment values were used for subsequent DIVAT biocollection (www.divat.fr) and stored in the Biologic analyses. The selection of the most discriminative genes to Resource Center of the Nantes University Hospital, F-44093, classify the three CD8 subsets was performed using sparse France (BRIF: BB-0033-00040). All donors were informed of partial least squares discriminant analysis (sPLS-DA)32 with the final use of their blood and signed a written-informed- the mixOmics package in R.33 A total of 100 repeats were used consent form. The University Hospital Ethical Committee to establish the sPLS-DA model. Two components within each and the Committee for the Protection of Patients from Bio- set of 100 genes were optimal to discriminate the three CD8 logic Risks approved the studies involving patients. Adults subsets. For gene expression representation, sample plots of who received kidney grafts from deceased donors and had a the sPLS-DA and clustering analysis results were generated in functional transplant at the first anniversary of transplanta- R version 3.5.1 using the ade434,35 and heatmap3 (https:// tion were included. Only patients without missing KTFS val- CRAN.R-project.org/package5heatmap3) packages, respec- ues were retained (i.e., patients for whom information about tively. The biologic significance of selected genes was assessed creatinemia at 3 and 12 months, proteinuria at 12 months, using High-Throughput GoMiner.36 Gene-ontology cate- recipient sex and age, number of previous transplants, last gories enriched with a false discovery rate of ,5% and con- donor creatinemia value, and number of rejection episodes taining at least five represented genes were selected. RNA-seq during the first year post-transplant were available). Finally, data were deposited in the Gene Expression Omnibus under the availability of frozen PBMCs at 1266 months was used the accession number GSE129356. to select the patients.

Statistical Analyses Biomarkers were dichotomized with respect to the median RESULTS value to avoid the log-linearity assumption (thus avoiding in- ﬂation of the type-1 error rate associated with the estimation Characteristics of the Cohort of the optimal cutoff value for predicting graft survival). Graft The demographic and clinical characteristics of the popula- survival was the primary outcome and was deﬁned as the time tion are shown in Table 1. The KTx underwent transplantation between the transplantation and the return to dialysis or pre- between July 2003 and January 2012. Among the 284 KTx alive emptive retransplantation (deaths were censored). Survival with a functional kidney graft at 1 year post-transplant (study curves were obtained using the Kaplan–Meier estimator.37 baseline), 57 returned to dialysis at the end of the follow-up, The raw and KTFS-adjusted associations were calculated using and 39 died. The median follow-up time was 8.3 years. At 5 Cox proportional hazards models.38 The Holm procedure was and 10 years, the graft survival rates were 91.5% (95% CI,

JASN 31: 876–891, 2020 TEMRA CD8+ TCells 879 CLINICAL RESEARCH www.jasn.org

Table 1. Description of the quantitative and qualitative characteristics at the for .5 years is a risk factor for graft dys- study baseline (i.e., 12 mo post-transplant; n5284) function.19 Here, we assessed the associ- Characteristics Mean6SD or Number (%) ation between the risk of kidney graft Recipient failure (a more stringent clinical crite- Age (yr) 51.07613.21 rion than that used in our previous Body mass index (kg/m2)24.2864.12 study) and the early monitoring of Cold ischemia time (h) 21.6068.11 CD8-related markers. Concurrent with Recipient creatinemia at 3 mo (mmol/L) 141.80649.92 the KTFS calculation, we monitored Recipient creatinemia at 12 mo (mmol/L) 139.05646.60 the frequency of CD8 T cell subsets us- Recipient proteinuria at 12 mo (g/24 h) 0.4160.82 ing the phenotypic markers CD45RA 6 1 1 KTFS 3.91 1.32 and CCR7 (naive, CD45RA CCR7 ; 1 2 First kidney transplant 253 (89.08) TEMRA, CD45RA CCR7 ;EM, Sex (male) 180 (63.38) 2 2 CD45RA CCR7 ; and central memory, Recurrent initial disease 81 (28.52) 2 1 CMV serology (positive) 148 (52.11) CD45RA CCR7 ), markers of differen- EBV serology (positive) 273 (96.81) tiation (CD27, CD28, and EOMES), the HCV serology (positive) 10 (3.52) expression of cytotoxic molecules .4 HLA mismatches (A1B1DR) 49 (17.25) (GZMB and PERF-1), and markers asso- $1 Acute rejection episode during the first year 30 (10.56) ciated with the secretion of proinflam- Diabetes 46 (16.20) matory cytokines (T-bet and CD57). Arterial hypertension 268 (94.37) After adjustment with the standard clin- Cardiovascular history 107 (37.68) ical metrics summarized in the KTFS, Cancer history 15 (5.28) the 1-year frequency of TEMRA CD8 Immunization with anti-HLA class I at day 0 53 (20.00) was associated with an increased risk Immunization with anti-HLA class II at day 0 47 (17.87) of graft failure (HR, 1.61; P50.083; Tacrolimus trough level (ng/ml) 8.6263.54 Donor Table 2). In contrast, an increased fre- Age (yr) 51.76616.36 quency of EM CD8 was associated with Last donor creatinemia (mmol/L) 87.61663.38 a reduced risk of graft failure (HR, 0.39; Male donor 169 (59.51) P50.0012; Table 2). The increase fre- Vascular cause of donor death 163 (57.39) quency of EM CD8 was associated with CMV, cytomegalovirus; EBV, Epstein–Barr virus; HCV, hepatitis C virus. a 10% higher frequency of anti- thymocyte globulin (ATG) as induction 88.2% to 94.9%) and 75.3% (95% CI, 69.2% to 82.1%), re- therapy (Supplemental Figure 3A). However, at the time of spectively (Figure 1A). For 89.1% of the KTx, this kidney analysis, the maintenance therapy was similar across KTx transplant was their first. The mean age of the recipients and (Supplemental Figure 3B). Finally, a higher frequency of donors was 51.1 years (range, 18–83 years) and 51.8 years chronic rejection in the KTx EMlow group as compared with (range, 6–82 years), respectively. the KTx EMhigh group was observed in for-cause kidney graft biopsy samples of the 57 patients with kidney graft failure (59% An Increase in the TEMRA/EM CD8 T Cell Ratio versus 42%, respectively; Supplemental Figure 3C). Collec- Identifies KTx at High Risk of Kidney-Graft Failure tively, our data suggest that the accumulation of TEMRA We first validated the predictive value of the KTFS in this new CD8 associated with a decrease in EM CD8 is not only associ- cohort of 284 KTx. In terms of discriminatory capacity, the ated with a higher risk of kidney graft failure but also that AUC at 8 years post-transplant was 0.75 (95% CI, 0.66 to 0.83) the accumulation of TEMRA CD8 is associated with a higher (Figure 1B), a prognostic value similar to that found in the frequency of immune-related rejection. original study (AUC, 0.78).21 Patients were stratified at 1 year The association between the TEMRA/EM CD8 proportion post-transplant according to the previously defined KTFS and kidney graft survival prompted us to hypothesize that the value21 into the low-risk group (KTFS#4.17) and the high- prognostic value of KTFS could be improved by combining risk group (KTFS.4.17), and we demonstrated that the graft the KTFS with the frequency of EM/TEMRA CD8 at 1 year survival rate was significantly different between the two post-transplant. As expected, a strong correlation was ob- groups (HR, 1.26; P,0.001; Figure 1B). At 8 years post- served between the percentages of TEMRA and EM CD8 transplant, the graft survival rates were 86.6% (95% CI, (P,0.001; Figure 1D), and we considered only the percentage 80.5% to 93.1%) and 61.6% (95% CI, 49.8% to 76.3%) for of the EM CD8 (the lowest P value in Table 2). Among the patients with a low risk and a high risk of graft failure, respec- patients at high risk of graft failure (KTFS.4.17; n585), the tively (Figure 1C). risk of graft failure was 2.3-fold (95% CI, 1.1 to 4.9) higher in We previously reported that an increase in highly differen- patients with an EM CD8 percentage ,36% (median of EM tiated TEMRA CD8 cells in patients with stable graft function CD8 in the cohort; n541) than in patients with a higher EM

880 JASN JASN 31: 876–891, 2020 www.jasn.org CLINICAL RESEARCH

A B 1.0 Survival 1.0 KTFS<4.17 95% CI KTFS>4.17 0.8 0.8

0.6 0.6

0.4 0.4

Number of at−risk patients Number of at−risk patients 0.2 284 278 265 253 237 216 181 137 101 72 34 13 7 0.2 198 194 191 185 175 161 140 105 80 59 29 10 5 86 84 74 68 62 55 41 32 21 13 5 3 2 Graft survival (deaths censored) Graft survival (deaths censored) 0.0 0.0 0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14 Time post transplant (years) Time post transplant (years) + - - + + - + - - + - CD- + + + - 1.0 + EOMES CD57 EOMES CD57 EOMES CD57 EOMES CD57 PERF PERF + high + + - - - + - PERF PERF CD27 + CD27 CD27 - - + - - +

0.8 EOMES TEMRA GZMb TBX21 TBX21 TBX21 TBX21 CD28 CD28 TBX21 GZMb CM TBX21 EM TBX21 CD28pCD27 GZMb GZMb NAIVE CD28 TBX21 TBX21 1 EOMES+ TEMRA 0.8 0.6 GZMb+PERF- TBX21+EOMES+ TBX21high 0.6 TBX21+CD57+ 0.4 TBX21+CD57- 0.4 Sensitivity CD28-CD27+ CD28-CD27- 0.2 TBX21+EOMES- - - 0.2 GZMb PERF 0 CM KTFS 8y (AUC=0.75) TBX21-EOMES- −0.2 KTFS 11y (AUC=0.79) EM 0.0 TBX21-CD57+ CD28+CD27- −0.4 0.0 0.2 0.4 0.6 0.8 1.0 GZMb-PERF+ GZMb+PERF+ −0.6 1 - Specificity NAIVE CD28+CD27+ −0.8 TBX21-EOMES+ - - E TBX21 CD57 −1 1.0 EM > 36% EM < 36% 0.8

0.6

0.4

0.2 Number of at−risk patients

Graft survival (deaths censored) 44 43 39 35 33 29 24 20 13 10 4 0.0 41 40 34 32 29 26 17 12 8 3 1 0 2 4 6 8 10 12 14 Time post transplant (years)

Figure 1. A low frequency of EM CD8 T cells at 1 year post-transplant identiﬁes patients with increased risk of graft failure. (A) Kidney graft survival in the overall cohort. Patients were included 12 months after transplantation, and the survival of their grafts was assessed using the Kaplan–Meier estimator. The number of patients at risk was calculated every year. (B) Graft survival curves according to the KTFS cutoff value. Patients were stratiﬁed according to the KTFS at 12 months post-transplant as low risk (KTFS#4.17) or high risk (KTFS.4.17), and graft survival was assessed using the Kaplan–Meier estimator. The number of patients at risk was calculated every year. (C) Prognostic value of the KTFS. Time-dependent ROC curves up to 8 and 11 years post-transplant related to the KTFS (n5284). (D) Correlogram of CD8-related markers measured 12 months post-transplant. Correlations were tested using Spearman rank

JASN 31: 876–891, 2020 TEMRA CD8+ TCells 881 CLINICAL RESEARCH www.jasn.org

CD8 percentage (n544). Therefore, we demonstrated that, supernatant from both TEMRA and EM CD8 (Figure 2B). in patients at high risk of graft failure, a 1-year decreased Finally, after donor-specific stimulation, vigorous proliferation percentage of EM CD8 T cells post-transplant is associated of naive, EM, and TEMRA CD8 was observed (Figure 2C). Of with the poorest prognosis. More precisely, in patients at note, syngeneic stimulation has previously been shown to fail to high risk of graft failure (KTFS.4.17), the 7-year survival elicit the activation of CD8 T cells12 and, in agreement with a rate was 62.4% (95% CI, 58.2% to 87.6%) and 82.6% (95% previous report,41 third-party stimulation elicits a strong upreg- CI, 75.5% to 97.0%) for patients with low or high EM CD8 ulation of CD25, associated with the expression of CD107a, only percentages, respectively (Figure 1E). Collectively, our data in the memory CD8 compartment (Supplemental Figure 5). Col- demonstrate that the measurement of CD8 memory subsets lectively, these findings demonstrate that TEMRA and EM CD8 at 1 year post-transplant improves the identification of pa- exhibit similar rapid and early functional responses to donor- tients who will later return to dialysis as a consequence of graft specific stimulation. failure. TEMRA CD8 Cells Exhibit an Innate-like Signature and TEMRA and EM CD8 Exhibit Similar Effector Functions Mediate ADCC upon CD16 Engagement upon Donor-Specific Stimulation The similar responses of TEMRA and EM CD8 from KTx after Given the low TCRVb repertoire diversity of TEMRA com- donor-specific stimulation prompted us to identify alternative pared with that of EM CD8,18,20 we hypothesized that TEMRA pathways of activation of TEMRA CD8 that could account CD8 are enriched in donor-specificreactiveCD8Tcells, for their pathogenic function leading to kidney graft failure. which could explain the inverse kidney graft outcomes be- Because we previously showed that the immune function of tween KTx stratified according to the TEMRA/EM CD8 TEMRA CD8 was similar between KTx and HV,20 eight sets of ratio. Donor and recipient PBMCs were collected from matched samples of naive, EM, and TEMRA CD8 isolated 24 living-donor KTx before and 1-year after transplant. We from the peripheral blood of HV were analyzed by RNA-seq first evaluated the consequence of kidney transplantation on to identify specific gene-expression profiles. Discriminatory the frequency and phenotype of CD8 subsets. The strong im- gene analysis (sPLS-DA; see the Methods section) revealed mune challenge induced by allogeneic kidney transplantation that CD8 subsets could be efficiently identified and grouped results in a decrease in naive CD8 T cells (31.67%63.13% (72.4% of the variance was explained by components 1 and 2; versus 23.60%62.54% before and 1-year after transplant, re- Figure 3A). The identification of CD8 subset-specificgene spectively; Supplemental Figure 4A) and an increase in signatures was confirmed by unsupervised clustering of TEMRA CD8 (24.69%63.76% versus 38.32%64.06% before CD8-subset transcriptomes in the discriminatory gene analy- and 1-year after transplant, respectively; Supplemental sis (Figure 3B). The top genes contributing to the identifica- Figure 4A). Native GZMB expression was restricted to tion of TEMRA CD8 were involved in cytotoxicity (GZMB, TEMRA CD8 (Supplemental Figure 4B) and, as expected, ex- GNLY,andPERF-1), transport of lysosomal enzymes pression of the TBX21 transcription factor and EOMES was (GNPTAB), cell surface receptor signal transmission with restricted to the memory (EM and TEMRA) CD8 cell com- key roles in the regulation of innate and adaptive immune partment (Supplemental Figure 4B). CD8 subsets were then responses (LYN ), and—interestingly —innate immuni- purified from living-donor KTx and stimulated with donor- ty–related functions (KIR3DL1, KIR3DL2, KLRD1, CD244 derived, T cell–depleted PBMCs. A strong upregulation of the or 2B4, CD300A,andFCGR3A or CD16) (Supplemental early activation marker CD69 was observed in naive and mem- Figure 6, Supplemental Table 2). Higher expression of CD16 ory (TEMRA and EM) CD8 after donor-specificstimulation was observed at the transcriptome level in TEMRA CD8 than (Figure 2A). However, the expression of the high-affinity in naive and EM CD8 (Figure 3C); this finding was confirmed 1 1 1 IL-2Ra chain, CD25, and the cytotoxic marker CD107a was via analysis of the phenotype of CD3 CD8 CD16 cells restricted to the memory CD8 subsets, and the magnitude of (61.1%65.1% TEMRA versus 8.4%62.6% EM; P,0.001; 1 1 1 CD25 and CD107a expression did not differ between EM and Figure 3D). Whereas most CD3 CD8 CD16 cells exhibited TEMRA CD8 (Figure 2A). This early and memory-restricted a TEMRA phenotype, a fraction of TEMRA CD8 expressed activation profile was confirmed by analysis of culture CD16 (11.8%63.9%; Figure 3E). To validate the expression of supernatant from donor-specific CD8 subsets (Figure 2B). CD16 by TEMRA CD8 and to gain additional insight into In addition, high levels of proinflammatory cytokines these CD8 subsets, we analyzed data from CyTOF performed (IFN-g,TNF-a, and IL-17A) and cytotoxic molecules (gran- on PBMCs from HV.26 Via unsupervised clustering (Pheno- ulysin, PERF-1, GZMA, and sFASL) were found in the Graph; see the Methods section) and t-distributed stochastic

correlation, and only signiﬁcant correlations are shown (P,0.01). (E) An EM CD8 T cell percentage of ,36% was associated with in- 2 2 creased risk of graft failure. Patients considered at high risk (KTFS.4.17) were stratiﬁed according to the median EM (CD45RA CCR7 ) CD8 T cell percentage 12 months post-transplant, and graft survival was assessed using the Kaplan–Meier estimator. The number of patients at risk was calculated every year.

882 JASN JASN 31: 876–891, 2020 www.jasn.org CLINICAL RESEARCH

Table 2. Bivariate Cox models of the association of CD8-related biomarkers with return to dialysis Marker Median HR P Value Corrected P Value EM 36.02 0.39 0.001 0.027 TEMRA 31.70 1.61 0.083 .0.99 Naive 14.65 0.85 0.56 .0.99 CM 3.72 1.13 0.67 .0.99 1 2 GZMb PERF-1 30.35 1.77 0.045 0.95 2 2 GZMb PERF-1 31.75 1.50 0.15 .0.99 1 1 GZMb PERF-1 22.10 0.78 0.37 .0.99 2 1 CD28 CD27 5.41 0.68 0.16 .0.99 2 2 CD28 CD27 9.92 1.48 0.17 .0.99 2 2 CD28 CD27 26.50 1.19 0.52 .0.99 1 1 CD28 CD27 42.90 0.94 0.84 .0.99 TBX21high 63.75 1.03 0.92 .0.99 1 EOMES 76.70 0.70 0.21 .0.99 1 2 TBX21 EOMES 7.47 0.76 0.33 .0.99 2 2 TBX21 EOMES 19.50 1.25 0.43 .0.99 2 1 TBX21 EOMES 20.55 1.22 0.48 .0.99 1 1 TBX21 EOMES 40.35 1.03 0.91 .0.99 2 2 TBX21 CD57 34.35 0.89 0.67 .0.99 1 2 TBX21 CD57 21.15 0.89 0.68 .0.99 2 1 TBX21 CD57 2.12 1.10 0.73 .0.99 1 1 TBX21 CD57 33.50 0.96 0.88 .0.99 HRs were adjusted for the KTFS. Corrected P values were obtained using the Holm method. CM, central memory. neighbor embedding to visualize high-dimension data with CD16 induces the activation of TEMRA CD8, as shown by single-cell resolution in two dimensions, we identified 23 clusters increases in proinflammatory cytokine secretion and cytotox- based on the expression of 24 markers (Supplemental Figure 7). icity compared with these parameters in EM CD8. Neither Major CD8 subsets (naive, central memory, EM, and TEMRA) IFN-g and TNF-a nor CD107a were found in naive CD8. were identified according to the expected expression of CD45RA Furthermore, a substantial increase in the effector response and CCR7 and were localized in neighboring locations on the of TEMRA CD8 was observed in the presence of IL-15 t-distributed stochastic neighbor embedding map. Eight subsets (Figure 3F). Irrelevant coated antibodies (anti-CD4 mAb, of TEMRA CD8 were identified (clusters 5, 9, 10, 16, 17, 18, 22, mouse IgG1; anti-HLA class I mAb, mouse IgG2a) failed to 1 2 2 and 23) with a shared phenotype (CD45RA CCR7 CD25 elicit a cytotoxic response by TEMRA, demonstrating the 2 2 1 1 1 1 CD62L CCR6 GZMb CD95 CD11a CD57 ). Most specificity of the CD16-triggered activation (Supplemental 1 2 TEMRA CD8 subsets were CCR5 CCR6 CCR4int. The TEMRA Figure 8C). Finally, we measured the ability of TEMRA CD8 subsets differed in their expression of costimulatory molecules to mediate ADCC by challenging CD8 T cells with Raji cells (CD28 and CD27), IL-7 receptor a-chain (CD127), and inhib- coated with anti-CD20 mAb (Figure 3G). Degranulation was itory molecules (BTLA, CTLA4, and LAG3). Interestingly, CD16 restricted to the TEMRA CD8 compartment (1.2%60.4% 1 expression was restricted to TEMRA subsets, in agreement with versus 8.3%63.2% CD107a for treatment with Raji the transcriptomic and flow cytometry data (Figure 3, A–E). Not cells versus treatment with Raji cells and anti-CD20 mAb, re- 1 only were TEMRA CD16 (clusters 5, 9, 10, 16, 17, 18 and 22) spectively; P50.031). Similar results were obtained for shorter 2 and TEMRA CD16 (cluster 23) subsets identified, but the ex- stimulation times (Supplemental Figure 8D). The degranula- pression of the activated innate-like marker CD56 also differed tion of TEMRA CD8 induced by ADCC was CD16 dependent among the TEMRA CD8 subsets (Supplemental Figure 7). because blocking aCD16 was sufficient to prevent the cyto- In summary, unsupervised analysis of large-scale CyTOF data toxic response of TEMRA CD8, whereas irrelevant Ig (anti- confirmed the existence of TEMRA CD8 subsets that could be HLA class I mAb) had no effect on the ADCC (Supplemental characterized according to CD16 expression, suggesting a poten- Figure 8E). These data demonstrate a unique innate-like sig- tial alternative pathway of TEMRA CD8 activation upon nature in TEMRA CD8 and the ability to activate TEMRA CD8 antibody-mediated Fc receptor engagement. by either the interaction between TCR and donor HLA/ To directly assess the functionality of CD16 in TEMRA peptide complex or by the activation of CD16 upon Ig ligation. CD8, purified naive, EM, and TEMRA CD8 were primed with plate-bound aCD16 (clone 3G8, mouse IgG1) for 4 hours, CD16 Engagement Selectively Activates TEMRA CD8 and the expression of the cytotoxic marker CD107a and pro- from KTx duction of IFN-g and TNF-a were assessed (Figure 3F, To investigate means by which chronic stimulation induced by Supplemental Figure 8, A–C). Short-term crosslinking of allogeneic kidney transplantation modified the innate-like

JASN 31: 876–891, 2020 TEMRA CD8+ TCells 883 CLINICAL RESEARCH www.jasn.org

A NAÏVE 2.89 0.93 100 *** 100 ***

85.4 ** ** + * * + 80 80

91.1 5.04 60 60

EM within CD69 +

4.18 19.4 within CD69 69.9 + 40 40

20 20

52.1 24.3 % of CD25 % of CD107a TEMRA 10.5 25.2 0 0 66.8 EM EM EM EM NAÏVE NAÏVE NAÏVE NAÏVE TEMRA TEMRA TEMRA TEMRA

37.0 27.3 M0 M12 M0 M12 CD69 CD107a CD8 CD25

B IL-2 IL-4 IL-6 IL-17A C * 10000 1000 * 10000 1000 * ****** **** * TEMRA 1000 100 1000 100 EM pg/mL 100 10 100

10 1 10 10 NAÏVE

IFNγ TNFα sFAS sFASL 100 100000 * 100000 *** 1000 1000 *** - * **** *** 10000 10000 80

1000 1000 100 100 60 pg/mL 100 100 within Annexin V 40 10 10 10 10 low

Granulysin PERF GzmA GzmB 20

10000 ** 10000 **** 10000 * 100000 % of CPD ** *** ** 0

10000 EM EM 1000 1000 1000 NAÏVE NAÏVE 1000 TEMRA TEMRA

pg/mL 100 100 100 M0 M12 100

10 10 10 10 EM EM EM EM EMRA EMRA EMRA EMRA NAÏVE NAÏVE NAÏVE NAÏVE T T T T

Figure 2. TEMRA and EM CD8 T cells from KTx exhibit similar potent effector responses upon donor-specific stimulation. CD8 T subsets (naive, EM, and TEMRA) were purified from KTx before (M0) and 12 months after transplantation (M12) and stimulated with donor-specific, T cell–depleted PBMCs for (A and B) 48 hours or (C) 5 days. (A) Frequencies of CD8 subsets expressing activation (CD69 and CD25) or degranulation (CD107a) markers. Representative flow data and the gating strategy are shown (n510). (B) Concentration of cytokines and cytotoxic molecules secreted by donor-specific stimulated CD8 subsets (n510). (C) Proliferation of CD8 subsets was 2 assessed according to the dilution of CPDeFluor450 signal within the subset of Annexin V cells. Representative flow data are shown (n55–6). The bars indicate the mean6SD, and each point represents a single KTx. The P values were calculated using nonparametric ANOVA (Kruskal–Wallis) with the Dunn multiple comparisons test. *P,0.05, **P,0.01, ***P,0.001, ****P,0.0001.

884 JASN JASN 31: 876–891, 2020 www.jasn.org CLINICAL RESEARCH

A B C CD16 Individual 8 10 EM NAÏVE EM TEMRA CD8 subset 6 5 4 0 2 log2 −5 0 Dim2 (30.1%) NAÏVE TEMRA −10 -2 -4 −10 0 10 Dim1 (42.3%) EM NAÏVE TEMRA

D E * NAÏVE 60 *** * 0.50 50 100 ****

+ 40 25 80 EM 0.96 20 60 15

40 % of CD16 10 TEMRA 5 14.0 20

% of CD8 subset 0 EM

0 CD16 NAÏVE

EM CD3 TEMRA NAÏVE TEMRA

F EM TEMRA 30 20 30 ** + ** ** +

+ 15

20 IFN γ 20 + ** 10 10 10 **

% of TNF α 5

** % of CD107a % of TNF α 0 0 0 CD16 -++ - CD16 -++ - CD16 - -+ + IL-15 --+ + IL-15 --+ + IL-15 - +- +

G NAÏVE EM TEMRA 25 0.33 0.28 0.15 Raji * + 20 15 Raji 0.86 2.03 19.6 10 + RTX

% of CD107a 5 CD107a CD3 0 RTX - + - + - + NAÏVE EM TEMRA

Figure 3. Selective activation of TEMRA CD8 T cells upon CD16 crosslinking. (A) PLS-DA sample plot and (B) heatmap showing scaled expression values of discriminating genes for CD8 T subsets (naive, EM, and TEMRA) purified from eight HV. (C) Expression of the 1 1 1 CD16 transcript by CD8 T cell subsets. (D) Phenotype of CD3 CD8 CD16 cells according to CD45RA and CCR7 expression (n513). 1 1 1 (E) Frequencies of CD16 cells among CD3 CD8 subsets. Representative flow data are shown (n513). (F) Frequencies of CD8 subsets secreting TNF-a, IFN-g, and CD107a after exposure to the indicated stimuli for 4 hours. Notably, naive CD8 T cells exhibited neither cytokine production nor degranulation. The bars indicate the mean6SEM of data from eight HV. (G) Expression of CD107a by CD8 subsets after 24 hours of coculture with Raji cells with or without rituximab (RTX). Representative flow data are shown (n55). The bars indicate the mean6SEM, and each point represents a single HV. The P values were calculated using nonparametric ANOVA

JASN 31: 876–891, 2020 TEMRA CD8+ TCells 885 CLINICAL RESEARCH www.jasn.org function of TEMRA CD8, IFN-g and TNF-a production was lower frequency of EM CD8 (i.e., a higher frequency of examined after short-term in vitro crosslinking of CD16 on TEMRACD8)thaninKTxwithahigherfrequencyof CD8 T cells purified from KTx. A significant and selective EM CD8. Consistent with our results obtained in KTx re- increase in the secretion of proinflammatory cytokines cruited .5 years after transplantation,19 these results showed by TEMRA CD8 from KTx relative to this secretion by that the modification of the CD8 compartment occurred at an EM CD8 was observed (Figure 4A). The agonist effect of earlier time point than originally thought and has a strong plate-bound aCD16 on TEMRA cells from KTx was further negative effect on kidney graft outcome. For the first time, enhanced upon exposure to IL-15 (Figure 4A). The cytotoxic we showed that KTx with a high risk of kidney graft failure activity of TEMRA CD8 from KTx was similarly triggered can be identified based on the combination of a clinical upon combined exposure to aCD16 and IL-15 stimulation metrics-based score with the monitoring of CD8 frequencies, (Figure 4A). Only TEMRA CD8 from KTx mediated ADCC a facile method in daily clinical practice. 1 (0.93%60.19% versus 4.15%61.77% CD107a for treat- A strong heterogeneity in the usage of the T cell compartment with Raji cells versus treatment with Raji cells ment is observed in KTx with stable graft function, despite and anti-CD20 mAb, respectively; P50.016; Figure 4B). We daily treatment with immunosuppressive drugs.16,18,19,42 The ultimately hypothesized that TEMRA CD8 could selectively results of this study further confirmed this finding, showing interact with HLA molecules, not only upon TCR-HLA inter- high variability in the frequency of EM and TEMRA CD8 action but also upon the binding of anti-HLA class II Ig to among KTx as early as 12 months post-transplant, a time point CD16 receptors expressed by TEMRA CD8. To test this hy- well into the induction therapy regimen. However, the in- pothesis, we incubated CD8 T cells with single-antigen HLA creased frequency of EM CD8 was associated with a 10% class II beads in the presence of serum from either immunized higher frequency of ATG as induction therapy and further KTx (with multiple HLA class II specificities) or from male studies are needed to decipher the susceptibility of CD8 sub- HVwithoutanyknownimmunization(Figure4Cand sets to anti-thymocyte globulin and the relative contribution Supplemental Figure 2). Exposure to immunized serum resul- of ATG to the lower kidney graft failure. In addition, this het- ted in selective binding of TEMRA CD8 T cells to HLA class II erogeneity demonstrates that, in some patients, the current molecules (percentage of CD8 bound to HLA class II mole- immunosuppressive drugs failed to prevent the increase of cules, 0.50%60.13% versus 0.17%60.04% for TEMRA and TEMRA CD8. The lack of TEMRA CD8-specific therapeutics EM, respectively; P50.004; Figure 4C). Finally, preincubation could be partially explained by the misunderstanding of of TEMRA CD8 with blocking aCD16 prevents the subsequent TEMRA CD8 functionality. This CD8 subset has long been interaction with DSA-coated HLA class II molecules considered a hallmark of immune senescence, and the elderly (P50.031; Figure 4D). Taken together, these results demon- and patients with chronic viral infections have historically strate that the activation of TEMRA CD8 from KTx could been the major populations of interest.43,44 However, expan- be achieved through either TCR or CD16 stimulation and sion of pathogenic TEMRA CD8 after allotransplantation has could foster the inflammatory response and promote kidney been documented in patients with autoimmune diseases graft failure. (multiple sclerosis,45 lupus,46,47 ANCA,48 or primary Sjögren syndrome49).19,50 TEMRA CD8 exhibit a potent inflammatory response when appropriately stimulated. Here, we demon- DISCUSSION strated that the response of TEMRA and EM CD8 does not differ after donor-specific stimulation; both memory CD8 The ability to stratify KTx according to the risk of graft failure populations favor the generation of a rapid inflammatory re- is a major challenge. Identifying patients with a high risk of sponse characterized by cytotoxic function (CD107a upregu- graft loss as early as possible may offer an early therapeutic lation and high levels of the cytotoxic molecules GZMA, window for intervention and, mainly, adaptation of the im- GNLY, and PERF-1) and the secretion of a wide spectrum of munosuppressive drug regimen. Clinical-based scoring sys- proinflammatory cytokines (IFN-g,TNF-a,andIL-17A) tems such as the KTFS21 form one approach to address this upon donor-specific stimulation. One hypothesis explaining challenge. However, patients’ survival prospects will greatly the negative effect of TEMRA CD8 on long-term graft survival benefit from biomarkers that combine the expectation of bio- could be related to an accumulation of preformed cytotoxic marker research (precision and sensitivity) with the cause molecules and proinflammatory cytokines (RNA-seq data of graft rejection. Here, we reported that monitoring EM/ and19) within TEMRA CD8 that could be released upon TEMRA CD8 in high-risk KTx (KTFS.4.17) enables the TCR stimulation via the direct pathway of allorecognition, identification of KTx with an immunologic risk of kidney graft which is maintained for years after transplantation and is neg- failure, because the graft survival rate was lower in KTx with a atively correlated with graft function.51 In addition to

(Kruskal–Wallis) with (D and E) the Dunn multiple comparisons test or (F and G) a Wilcoxon matched-pairs signed-rank test. *P,0.05, **P,0.01, ***P,0.001. Dim, dimension.

886 JASN JASN 31: 876–891, 2020 www.jasn.org CLINICAL RESEARCH

A 15 5 ** 15 **

+ 4 + + **

10 IFN 10 + 3

2 ** 5 ** ** 5 % of TNF 1 % of CD107 % of TNF

0 0 0 CD16 -++ - CD16 -++ - CD16 --++ IL-15 --+ + IL-15 --+ + IL-15 --+ +

NAÏVE EM TEMRA B 25 Raji 0.52 1.21 1.64 *

+ 20

Raji 1.23 2.51 12.0 10 + RTX % of CD107a 5

CD107a 0 CD3 RTX - + - + - + NAÏVE EM TEMRA

C EM TEMRA D Serum 1.5 ** 0.5 control * 0.095% 0.29% 0.4 1.0 0.3 Serum 0.2 Immunized 0.5 % of rosette KTx % of rosette 0.13% 0.95% 0.1

Beads 710/50 640nm 0.0 0.0 Beads EM TEMRA 660/20 561nm Control aCD16

Figure 4. Activation of the CD16 pathway selectively induces the proinflammatory response of TEMRA CD8 T cells from KTx. (A) Frequencies of CD8 subsets secreting TNF-a,IFN-g, and CD107a after exposure to the indicated stimuli for 4 hours. Notably, naive CD8 T cells exhibited neither cytokine production nor degranulation. The bars indicate the mean6SEM of data from seven KTx. (B) Expression of CD107a by CD8 subsets after 24 hours of coculture with Raji cells with or without rituximab (RTX). Representative flow 1 data are shown (n56). The bars indicate the mean6SEM, and each point represents a single KTx. (C and D) Frequencies of CD3 CD8 subsets bound to (C) single-antigen HLA class II beads coated with serum from nonimmunized male individuals (serum control) or with serum from immunized KTx (n59) or (D) before and after preincubation of TEMRA CD8 with aCD16 (n56). The P values were calculated using nonparametric ANOVA (Kruskal–Wallis) with (A) the Dunn multiple comparisons test or (B) a Wilcoxon matched-pairs signed-rank test. *P,0.05, **P,0.01. identifying the potent TCR-dependent response of TEMRA functional CD16 upon activation with IL-2 or with Ag and CD8, our report clarifies a unique feature restricted to Il-2,52 as shown by the ability of self-specific CD8 to mediate TEMRA CD8 that could account for the poor clinical out- ADCC independent of the TCR.52 The level of CD16 expres- comes in KTx with a high frequency of TEMRA CD8. TEMRA sion by NK cells is negatively regulated by CD3z,53 asubunit CD8 express a transcriptomic signature associated with the component of the FcgR CD16 and TCR/CD3 complex.54 For innate-like population, including the expression of CD16 example, induced expression of CD3z in murine NK cells en- (FcgRIIIa). The engagement of CD16 results in the selective hances the formation of CD3z/FcRg heterodimers, prevents activation of TEMRA CD8, characterized by proinflammatory association with CD16,52,53 and decreases their ADCC func- cytokine secretion and cytotoxic responses. Self-specificacti- tion, suggesting that a delicate balance of CD16/CD3z is re- vated CD44hi CD8 mouse T cells were reported to express quired to limit the expression of CD16. Our RNA-seq results

JASN 31: 876–891, 2020 TEMRA CD8+ TCells 887 CLINICAL RESEARCH www.jasn.org revealed that TEMRA CD8 exhibited upregulated expression activation of TEMRA CD8 and, therefore, could lead to an of the genes encoding the signaling adaptor FceRIg,anITAM- improvement in long-term graft outcomes. bearing adaptor molecule known to regulate cell surface In summary, we hypothesize that TEMRA CD8 migrate to CD16 expression and function, and its downstream signaling the graft, where donor HLA promotes their activation upon molecules 1-phosphatidylinositol-4,5-bisphosphate phos- recognition by TCRs or engagement of the CD16 pathway phodiesterase g-2 (PLCG2) and zinc finger and BTB by anti-HLA antibodies, thus favoring the induction of a domain–containing protein 16 encoding the promyelocytic sustained inflammatory microenvironment, including endo- leukemia zinc finger (ZBTB16), which interact with the thelial activation.20 This original mechanism of activation FceRIg promoter55). We also found decreased expression of suggests that TEMRA CD8 are involved in cellular and hu- the transcription factor Bcl11b, which is reported to protect moral rejection of kidney grafts and that KTx will benefit from T-cell identity.56 The regulation of CD16 expression on the monitoring of CD8 T cell subsets and the development of TEMRA CD8, the identification of signaling events leading therapeutics specifically targeting TEMRA CD8. to CD16 expression, and the interconnection between the TCR signaling complex and CD16/FcgR require further in- vestigation. However, our data support the concept of CD8 T-cell plasticity with the acquisition of innate-like func- ACKNOWLEDGMENTS tions by TEMRA CD8. With a dual activation mechanism that relies on TCR and FcgR signals, TEMRA CD8 are likely to We would like to thank Dr. Bertram Bengsch and Dr. E. John Wherry promote and sustain an inflammatory environment leading for kindly sharing the CyTOF data and Samuel Granjeaud for his to kidney graft failure. invaluable help with the R-based cytometric analysis. Few reports have highlighted the expression of FcgRby The Biological Resource Center of the Nantes University Hospital, human T cells. FcgRIIIa (CD16) expression was reported on F-44093, France (BRIF: BB-0033-00040) guarantees the quality of the human CD4 T cells from HV and patients with SLE, and its biologic samples. binding to immune complexes induces high secretion of Dr. Jacquemont and Ms. Tilly designed the experiments, per- IFN-g.57 Memory CD8 T cells generated by viral or bacterial formed the experiments, and analyzed the data. Dr. Yap and infection were shown to selectively express FcgRIIB and, Dr. Doan-Ngoc performed the experiments and analyzed the data. upon engagement, contribute to inhibiting the cytotoxicity Dr. Danger analyzed the transcriptomic data. Mr. Guerif and of memory CD8 T cells and their expansion after homologous Dr. Giral assisted with human sample collection and processing, challenge.58 In addition, interaction between T cells and with patient consent, and they obtained ethical approval for human FcgR-expressing antigen-presenting cells was also shown to studies. Dr. Delbos and Mr. Martinet provided critical reagents. result in the acquisition of FcgR by T cells via trogocytosis.59 Dr. Foucher contributed to biostatistical data analyses. Dr. Degauque, Despite the scarcity of reports of FcgR expression by human Dr. Brouard, Dr. Giral, Dr. Foucher, and Dr. Jacquemont designed and T cells, the expression of NK-related markers such as KIRs and supervised the study. Dr. Degauque performed the experiments, ana- NKG2A by CD8ab cells has long been reported.60,61 Increas- lyzed the data, and wrote the manuscript with input from all authors. ing evidence is available regarding both the expression of innate-associated markers by naive62 and memory63–65 CD8 T cells and the identification of innate-like CD8 T cells. 1 DISCLOSURES KIR/NKG2A CD8 T cells—identified in the blood of HV,64 64 in cord blood, and in patients with chronic myeloid leuke- None. mia65—rapidly produce IFN-g in response to IL-12 and IL-18 stimulation and exhibit antigen-independent cytotoxic function. Exposure to IL-15 not only maintains the expression of 1 FUNDING NKp30 in NKp30 CD8 T cells but also promotes the differ- 1 entiation of NKp30 CD8 T cells with concomitant acquisi- Dr. Yap is supported by a Fondation ProGreffe international fellowship tion of other NK receptors, high expression of T-bet, and low grant. Dr. Jacquemont received a Société Française de Transplantation grant. expression of the transcription factor Bcl11b and exhaustion This work was funded by ITMO Santé Publique grant A13053NS, supported markers (PD-1, CTLA-4, and Lag-3).62 We and others have by Agence Nationale de la Recherche (ANR; French National Research Agency) grant ANR-11-JSV1-0008-01, and supported in part by Réseau shown that IL-15 regulates the function and homeostasis of Thématique de Recherche et de Soins (RTRS) Fondation de Coopération 20,66,67 TEMRA CD8 T cells. Signaling via IL-2 and IL-15 in- Scientifique CENTAURE grants PAC8 and PAC9. This work was performed duces the loss of CD28,66,68 a phenotypic characteristic as a part of the IHU-CESTI project, which received French government finan- of TEMRA CD842,69; moreover, the shared IL-2 and IL-15 re- cial support managed by the ANR via the “Investment into the Future” pro- ceptor b-chain CD122 was shown to be critical for gram ANR-10-IBHU-005. The IHU-CESTI project is also supported by the 70 Nantes Metropolis and the Conseil Régional des Pays de la Loire. This work costimulation-independent T cell alloreactivity. Selective in- was also supported by the FP7 VISICORT project, which has received funding hibition of IL-15 signaling in TEMRA CD8 could thus be an from the European Union’s Seventh Framework Programme for research, appealing therapeutic strategy to limit the expansion and technological development, and demonstration under grant agreement

888 JASN JASN 31: 876–891, 2020 www.jasn.org CLINICAL RESEARCH

602470. This work was performed as a part of the LabEX IGO program sup- 7. Hidalgo LG, Sis B, Sellarés J, Campbell PM, Mengel M, Einecke G, ported by the ANR via the Investment into the Future program ANR-11- et al.: NK cell transcripts and NK cells in kidney biopsies from pa- LABX-0016-01. This work was supported in the context of the ANR project tients with donor-specific antibodies: Evidence for NK cell in- BIKET (ANR-17-CE17-0008). volvement in antibody-mediated rejection. Am J Transplant 10: 1812–1822, 2010 8. Hidalgo LG, Sellarés J, Sis B, Mengel M, Chang J, Halloran PF: Inter- preting NK cell transcripts versus T cell transcripts in renal transplant SUPPLEMENTAL MATERIAL biopsies. Am J Transplant 12: 1180–1191, 2012 9. Loupy A, Lefaucheur C, Vernerey D, Chang J, Hidalgo LG, Beuscart T, This article contains the following supplemental material online at et al.: Molecular microscope strategy to improve risk stratification in http://jasn.asnjournals.org/lookup/suppl/doi:10.1681/ASN.2019080847/-/ early antibody-mediated kidney allograft rejection. J Am Soc Nephrol 25: 2267–2277, 2014 DCSupplemental. 10. Lefaucheur C, Viglietti D, Hidalgo LG, Ratner LE, Bagnasco SM, Batal I, Supplemental Table 1. Gene ontology (GO) categories of CD8 et al.: Complement-activating Anti-HLA antibodies in kidney trans- subset gene signatures identified by sparse partial least squares dis- plantation: Allograft gene expression profiling and response to treat- criminant analysis (sPLS-DA). ment. JAmSocNephrol29: 620–635, 2018 11. Poggio ED, Clemente M, Riley J, Roddy M, Greenspan NS, Dejelo C, Supplemental Table 2. List of monoclonal antibodies used. et al.: Alloreactivity in renal transplant recipients with and without Supplemental Figure 1. The NK gene signature in kidney biopsies chronic allograft nephropathy. JAmSocNephrol15: 1952–1960, 2004 from KTx with ABMR is not restricted to the NK compartment. 12. van de Berg PJ, Yong SL, Koch SD, Lardy N, van Donselaar-van der Pant Supplemental Figure 2. Gating strategy to analyze the binding of KA, Florquin S, et al.: Characteristics of alloreactive T cells measured CD8 subsets with serum-coated single-antigen HLA class II beads. before renal transplantation. Clin Exp Immunol 168: 241 –250, 2012 13. Ashton-Chess J, Dugast E, Colvin RB, Giral M, Foucher Y, Moreau A, Supplemental Figure 3. Kidney biopsies from patients with graft et al.: Regulatory, effector, and cytotoxic T cell profiles in long-term failure and low frequency of EM CD8 display higher rate of ABMR kidney transplant patients. JAmSocNephrol20: 1113–1122, 2009 and TCMR. 14. Homs S, Mansour H, Desvaux D, Diet C, Hazan M, Buchler M, et al.: Supplemental Figure 4. TEMRA CD8 T cells from living donor Predominant Th1 and cytotoxic phenotype in biopsies from renal KTx and deceased donor KTx exhibit similar phenotypes. transplant recipients with transplant glomerulopathy. Am J Transplant 9: 1230–1236, 2009 Supplemental Figure 5. TEMRA and EM CD8 T cells from KTx 15. Obata F, Yoshida K, Ohkubo M, Ikeda Y, Taoka Y, Takeuchi Y, et al.: exhibit similar potent effector responses upon third party Contribution of CD41 and CD81 T cells and interferon-gamma to the stimulation. progress of chronic rejection of kidney allografts: The Th1 response Supplemental Figure 6. Unique gene signatures discriminate CD8 mediates both acute and chronic rejection. Transpl Immunol 14: 21–25, subsets via PLS-DA. 2005 16. Baeten D, Louis S, Braud C, Braudeau C, Ballet C, Moizant F, et al.: Supplemental Figure 7. Phenotype of CD16-expressing CD8 Phenotypically and functionally distinct CD81 lymphocyte populations T cells using high-dimensional data analysis. in long-term drug-free tolerance and chronic rejection in human kidney Supplemental Figure 8. Early and selective activation of TEMRA graft recipients. JAmSocNephrol17: 294–304, 2006 CD8 T cells upon CD16 crosslinking. 17. Brouard S, Le Bars A, Dufay A, Gosselin M, Foucher Y, Guillet M, et al.: Identification of a gene expression profile associated with operational tolerance among a selected group of stable kidney transplant patients. Transpl Int 24: 536–547, 2011 REFERENCES 18. Miqueu P, Degauque N, Guillet M, Giral M, Ruiz C, Pallier A, et al.: Analysis of the peripheral T-cell repertoire in kidney transplant patients. 1. Loupy A, Lefaucheur C, Vernerey D, Prugger C, Duong van Huyen JP, Eur J Immunol 40: 3280–3290, 2010 Mooney N, et al.: Complement-binding anti-HLA antibodies and 19. Yap M, Boeffard F, Clave E, Pallier A, Danger R, Giral M, et al.: Ex- kidney-allograft survival. N Engl J Med 369: 1215–1226, 2013 pansion of highly differentiated cytotoxic terminally differentiated ef- 2. Lefaucheur C, Loupy A, Vernerey D, Duong-Van-Huyen J-P, Suberbielle fector memory CD81 T cells in a subset of clinically stable kidney C, Anglicheau D, et al.: Antibody-mediated vascular rejection of kidney transplant recipients: A potential marker for late graft dysfunction. JAm allografts: A population-based study. Lancet 381: 313–319, 2013 Soc Nephrol 25: 1856–1868, 2014 3. Roufosse C, Simmonds N, Clahsen-van Groningen M, Haas M, 20. Tilly G, Doan-Ngoc T-M, Yap M, Caristan A, Jacquemont L, Danger R, Henriksen KJ, Horsfield C, et al.: A 2018 reference guide to the banff et al.: IL-15 harnesses pro-inflammatory function of TEMRA CD8 in classification of renal allograft pathology. Transplantation 102: kidney-transplant recipients. Front Immunol 8: 778, 2017 1795–1814, 2018 21. Foucher Y, Daguin P, Akl A, Kessler M, Ladrière M, Legendre C, et al.: A 4. Aubert O, Loupy A, Hidalgo L, Duong van Huyen J-P, Higgins S, clinical scoring system highly predictive of long-term kidney graft sur- Viglietti D, et al.: Antibody-mediated rejection due to preexisting ver- vival. Kidney Int 78: 1288–1294, 2010 sus de novo donor-specific antibodies in kidney allograft recipients. 22. Kaboré R, Haller MC, Harambat J, Heinze G, Leffondré K: Risk pre- JAmSocNephrol28: 1912–1923, 2017 diction models for graft failure in kidney transplantation: A systematic 5. Lefaucheur C, Viglietti D, Bentlejewski C, Duong van Huyen J-P, review. Nephrol Dial Transplant 32[Suppl 2]: ii68–ii76, 2017 Vernerey D, Aubert O, et al.: IgG donor-specific anti-human HLA anti- 23. Foucher Y, Meurette A, Daguin P, Bonnaud-Antignac A, Hardouin J-B, body subclasses and kidney allograft antibody-mediated injury. JAm Chailan S, et al.: A personalized follow-up of kidney transplant recipi- Soc Nephrol 27: 293–304, 2016 ents using video conferencing based on a 1-year scoring system pre- 6. Bouquegneau A, Loheac C, Aubert O, Bouatou Y, Viglietti D, Empana dictive of long term graft failure (TELEGRAFT study): Protocol for a J-P, et al.: Complement-activating donor-specific anti-HLA antibodies randomized controlled trial. BMC Nephrol 16: 6, 2015 and solid organ transplant survival: A systematic review and meta- 24. Braun WE, Schold JD: Transplantation: Strength in numbers-predicting analysis. PLoS Med 15: e1002572, 2018 long-term transplant outcomes. Nat Rev Nephrol 7: 135–136, 2011

JASN 31: 876–891, 2020 TEMRA CD8+ TCells 889 CLINICAL RESEARCH www.jasn.org

25. Moore J, He X, Liu X, Shabir S, Ball S, Cockwell P, et al.: Mortality 46. McKinney EF, Lyons PA, Carr EJ, Hollis JL, Jayne DR, Willcocks LC, prediction after kidney transplantation: Comparative clinical use of 7 et al.: A CD81 T cell transcription signature predicts prognosis in au- comorbidity indices. Exp Clin Transplant 9: 32–41, 2011 toimmune disease. Nat Med 16: 586–591, 2010 26. Bengsch B, Ohtani T, Khan O, Setty M, Manne S, O’Brien S, et al.: 47. McKinney EF, Lee JC, Jayne DR, Lyons PA, Smith KG: T-cell exhaustion, Epigenomic-guided mass cytometry profiling reveals disease-specific co-stimulation and clinical outcome in autoimmunity and infection. features of exhausted CD8 T cells. Immunity 48: 1029–1045.e5, 2018 Nature 523: 612–616, 2015 27. Becher B, Schlitzer A, Chen J, Mair F, Sumatoh HR, Teng KW, et al.: 48. Néel A, Bucchia M, Néel M, Tilly G, Caristan A, Yap M, et al.: Damp- High-dimensional analysis of the murine myeloid cell system. Nat Im- ening of CD81 T cell response by B cell depletion therapy in anti- munol 15: 1181–1189, 2014 neutrophil cytoplasmic antibody–associated vasculitis. Arthritis 28. Chen H, Lau MC, Wong MT, Newell EW, Poidinger M, Chen J: Cytofkit: Rheumatol 71[4]: 641–650, 2019 A bioconductor package for an integrated mass cytometry data analysis 49. Tasaki S, Suzuki K, Nishikawa A, Kassai Y, Takiguchi M, Kurisu R, et al.: pipeline. PLOS Comput Biol 12: e1005112, 2016 Multiomic disease signatures converge to cytotoxic CD8 T cells in 29. Wong MT, Chen J, Narayanan S, Lin W, Anicete R, Kiaang HT, et al.: primary Sjögren’ssyndrome.Ann Rheum Dis 76: 1458–1466, 2017 Mapping the diversity of follicular helper T cells in human blood and 50. Reinke S, Geissler S, Taylor WR, Schmidt-Bleek K, Juelke K, tonsils using high-dimensional mass cytometry analysis. Cell Reports Schwachmeyer V, et al.: Terminally differentiated CD8⁺ T cells nega- 11: 1822–1833, 2015 tively affect bone regeneration in humans [published correction ap- 30. Trombetta JJ, Gennert D, Lu D, Satija R, Shalek AK, Regev A: Prepa- pears in Sci Transl Med 5: 187er4, 2013]. Sci Transl Med 5: 177ra36, ration of single-cell RNA-Seq libraries for next generation sequencing. 2013 Curr Protoc Mol Biol 107: 1–17, 2014 51. Bestard O, Nickel P, Cruzado JM, Schoenemann C, Boenisch O, Sefrin 31. Trapnell C, Roberts A, Goff L, Pertea G, Kim D, Kelley DR, et al.: Dif- A, et al.: Circulating alloreactive T cells correlate with graft function in ferential gene and transcript expression analysis of RNA-seq experi- longstanding renal transplant recipients. JAmSocNephrol19: ments with TopHat and Cufflinks. Nat Protoc 7: 562–578, 2012 1419–1429, 2008 32. Lê Cao KA, Boitard S, Besse P: Sparse PLS discriminant analysis: Bi- 52. Dhanji S, Tse K, Teh H-S: The low affinity Fc receptor for IgG functions ologically relevant feature selection and graphical displays for multi- as an effective cytolytic receptor for self-specificCD8Tcells.JImmunol class problems. BMC Bioinformatics 12: 253, 2011 174: 1253–1258, 2005 33. Rohart F, Gautier B, Singh A, Lê Cao KA: mixOmics: An R package for 53. Arase H, Suenaga T, Arase N, Kimura Y, Ito K, Shiina R, et al: Negative ‘omics feature selection and multiple data integration. PLOS Comput regulation of expression and function of Fc gamma RIII by CD3 zeta in Biol 13: e1005752, 2017 murine NK cells. JImmunol166: 21–25, 2001 34. Bougeard S, Dray S: Supervised multiblock analysis in Rwith the 54. Kurosaki T, Gander I, Ravetch JV: A subunit common to an IgG Fc re- ade4Package. JStatSoftw86: 1–17, 2018 ceptor and the T-cell receptor mediates assembly through different 35. Dray S, Dufour A-B: The ade4 Package: Implementing the duality dia- interactions. Proc Natl Acad Sci U S A 88: 3837–3841, 1991 gram for ecologists. J Stat Softw 22: 1–20, 2007 55. Schlums H, Cichocki F, Tesi B, Theorell J, Béziat V, Holmes TD, et al.: 36. Zeeberg BR, Qin H, Narasimhan S, Sunshine M, Cao H, Kane DW, et al.: Cytomegalovirus infection drives adaptive epigenetic diversification of High-Throughput GoMiner, an ‘industrial-strength’ integrative gene NK cells with altered signaling and effector function. Immunity 42: ontology tool for interpretation of multiple-microarray experiments, 443–456, 2015 with application to studies of Common Variable Immune Deficiency 56. Li P, Burke S, Wang J, Chen X, Ortiz M, Lee S-C, et al.: Reprogramming (CVID). BMC Bioinformatics 6: 168, 2005 of T cells to natural killer-like cells upon Bcl11b deletion. Science 329: 37. Kaplan E, Meier P: Nonparametric estimation from incomplete obser- 85–89, 2010 vations. JAmStatAssoc53: 457–481, 1958 57. Chauhan AK, Chen C, Moore TL, DiPaolo RJ: Induced expression of 38. Cox D: Regression models and life-tables. J Roy Stat Soc Series B Stat FcgRIIIa (CD16a) on CD41 T cells triggers generation of IFN-ghigh Methodol 34: 187–220, 1972 subset. JBiolChem290: 5127–5140, 2015 39. Blanche P, Dartigues J-F, Jacqmin-Gadda H: Review and comparison of 58. Starbeck-Miller GR, Badovinac VP, Barber DL, Harty JT: Cutting edge: ROC curve estimators for a time-dependent outcome with marker- Expression of FcgRIIB tempers memory CD8 T cell function in vivo. dependent censoring. Biom J 55: 687 –704, 2013 JImmunol192: 35–39, 2014 40. Benjamini Y, Krieger AM, Yekutieli D: Adaptive linear step-up proce- 59. Hudrisier D, Clemenceau B, Balor S, Daubeuf S, Magdeleine E, Daëron dures that control the false discovery rate. Biometrika 93: 491–507, M, et al.: Ligand binding but undetected functional response of FcR 2006 after their capture by T cells via trogocytosis. JImmunol183: 41. Fischer M, Leyking S, Schäfer M, Elsäßer J, Janssen M, Mihm J, et al.: 6102–6113, 2009 Donor-specific alloreactive T cells can be quantified from whole blood, 60. Mingari MC, Moretta A, Moretta L: Regulation of KIR expression in and may predict cellular rejection after renal transplantation. Eur human T cells: A safety mechanism that may impair protective T-cell J Immunol 47: 1220–1231, 2017 responses. Immunol Today 19: 153–157, 1998 42. Yap M, Tilly G, Giral M, Brouard S, Degauque N: Benefits of using 61. Mingari MC, Vitale C, Cambiaggi A, Schiavetti F, Melioli G, Ferrini S, CD45RA and CD28 to investigate CD8 subsets in kidney transplant et al.: Cytolytic T lymphocytes displaying natural killer (NK)-like ac- recipients. Am J Transplant 16: 999–1006, 2016 tivity: Expression of NK-related functional receptors for HLA class I 43. Wertheimer AM, Bennett MS, Park B, Uhrlaub JL, Martinez C, Pulko V, molecules (p58 and CD94) and inhibitory effect on the TCR-mediated et al.: Aging and cytomegalovirus infection differentially and jointly target cell lysis or lymphokine production. Int Immunol 7: 697–703, affect distinct circulating T cell subsets in humans. JImmunol192: 1995 2143–2155, 2014 62. Correia MP, Stojanovic A, Bauer K, Juraeva D, Tykocinski L-O, Lorenz 1 1 1 44. Khan N, Shariff N, Cobbold M, Bruton R, Ainsworth JA, Sinclair AJ, H-M, et al.: Distinct human circulating NKp30 Fc«RIg CD8 Tcell et al.: Cytomegalovirus seropositivity drives the CD8 T cell repertoire population exhibiting high natural killer-like antitumor potential. Proc toward greater clonality in healthy elderly individuals. J Immunol 169: Natl Acad Sci U S A 115: E5980–E5989, 2018 1984–1992, 2002 63. Kim J, Chang D-Y, Lee HW, Lee H, Kim JH, Sung PS, et al.: Innate-like 1 45. Salou M, Garcia A, Michel L, Gainche-Salmon A, Loussouarn D, cytotoxic function of bystander-activated CD8 T cells is associated Nicol B, et al.: Expanded CD8 T-cell sharing between periphery with liver injury in acute hepatitis A. Immunity 48: 161–173.e5, 2018 and CNS in multiple sclerosis. Ann Clin Transl Neurol 2: 609–622, 64. Jacomet F, Cayssials E, Basbous S, Levescot A, Piccirilli N, Desmier 2015 D, et al.: Evidence for eomesodermin-expressing innate-like CD8(1)

890 JASN JASN 31: 876–891, 2020 www.jasn.org CLINICAL RESEARCH

KIR/NKG2A(1) T cells in human adults and cord blood samples. Eur 68. MiyagawaF,TagayaY,KimBS,PatelHJ,IshidaK,OhtekiT,etal.:IL-15 J Immunol 45: 1926–1933, 2015 serves as a costimulator in determining the activity of autoreactive CD8 65. Jacomet F, Cayssials E, Barbarin A, Desmier D, Basbous S, Lefèvre L, T cells in an experimental mouse model of graft-versus-host-like dis- et al.: The Hypothesis of the human iNKT/Innate CD8(1) T-cell axis ease. JImmunol181: 1109–1119, 2008 applied to cancer: Evidence for a deﬁciency in chronic myeloid leuke- 69. Appay V, Dunbar PR, Callan M, Klenerman P, Gillespie GM, mia. Front Immunol 7: 688, 2017 Papagno L, et al.: Memory CD81 T cells vary in differentiation 66. Chiu WK, Fann M, Weng NP: Generation and growth of CD28nullCD81 phenotype in different persistent virus infections. Nat Med 8: memory T cells mediated by IL-15 and its induced cytokines. J Immunol 379–385, 2002 177: 7802–7810, 2006 70. Mathews DV, Dong Y, Higginbotham LB, Kim SC, Breeden CP, 67. Setoguchi R: IL-15 boosts the function and migration of human termi- Stobert EA, et al.: CD122 signaling in CD81 memory T cells drives nally differentiated CD81 T cells by inducing a unique gene signature. costimulation-independent rejection. JClinInvest128: 4557–4572, Int Immunol 28: 293–305, 2016 2018

AFFILIATIONS

1Université de Nantes, Inserm, Centre de Recherche en Transplantation et Immunologie (CRTI), UMR 1064, Nantes, France; 2CHU Nantes, Université de Nantes, ITUN, Nantes, France; 3Etablissement Français du Sang, Nantes, France; and 4INSERM, Université de Nantes, methodS in Patient-centered outcomes and HEalth ResEarch (SPHERE), UMR1246, Nantes, France

JASN 31: 876–891, 2020 TEMRA CD8+ TCells 891 Supplemental Table of Contents

Supplementary Table 1. Gene ontology (GO) categories of CD8 subset gene signatures identified by sparse partial least squares discriminant analysis (sPLS-DA).

Supplementary Table 2. List of monoclonal antibodies used.

Supplementary Figure 1. The NK gene signature in kidney biopsies from KTx with ABMR is not restricted to the NK compartment.

Supplementary Figure 2. Gating strategy to analyze the binding of CD8 subsets with serum- coated single-antigen HLA class II beads.

Supplementary Figure 3. Kidney biopsies from patients with graft failure and low frequency of EM CD8 display higher rate of ABMR and TCMR

Supplementary Figure 4. TEMRA CD8 T cells from living donor KTx and deceased donor

KTx exhibit similar phenotypes.

Supplementary Figure 5. TEMRA and EM CD8 T cells from KTx exhibit similar potent effector responses upon 3rd party stimulation.

Supplementary Figure 6. Unique gene signatures discriminate CD8 subsets via PLS-DA.

Supplementary Figure 7. Phenotype of CD16-expressing CD8 T cells using high-dimensional data analysis.

Supplementary Figure 8. Early and selective activation of TEMRA CD8 T cells upon CD16 crosslinking.

A B 1.0 1.0 MYBL1 0.5 0.5 EM GNLY 15.0 0.0 CX3CR1 0.0 KLRF1 14.5 FCGR3A 14.0 FGFBP2

Dim2 (6.7%) −0.5 Dim2 (6.7%) SH2D1B 13.5 −0.5 −1.0 TEMRA NAÏVE −1.5 −1.0 −4 −2 0 2 −1.0 −0.5 0.0 0.5 1.0 Dim1 (86.4%) Dim1 (86.4%)

C NAÏVE EM TEMRA GNLY CX3CR1 FCGR3A FGFBP2 KLRF1 SH2D1B MYBL1

Supplementary Figure 1 250K 5 5 Beads CD8 + Beads 10 10 Rosette_TEMRA

200K

4 4 10 10

150K

SSC EM TEMRA 3 3 100K 10 10 CCR7

0 710/50 640nm 0 50K Lymphocyte

3 3 -10 -10 0 4 5 3 3 4 5 0 50K 100K 150K 200K 250K 0 10 10 -10 0 10 10 10 FSC CD45RA 660/20 561nm

5 5 10 10 Rosette_EM

4 4 10 10

3 3

CCR7 10 10

0 710/50 640nm 0

3 3 -10 EM TEMRA -10 4 5 3 3 4 5 0 10 10 -10 0 10 10 10 CD45RA 660/20 561nm

Supplementary Figure 2 A Induction therapy EM low EM high

0.70% 2.11%

78.87% 64.08% No Induction Basiliximab 20.42% 33.80% Thymoglobulin

B Maintenance therapy Steroids mTOR inh. MMF Calcineurin inh. 100

60 No 40 Yes % of patients 20

0 EM low EM high EM low EM high EM low EM high EM low EM high

C CAMR 5% CAMR 38% 33% 19% 14% 19% 30% 3% EM low high EM 14% 19% 8% 34% Immune 9% Immune 8% 13% Rejection 5% Rejection 59% 42%

CAMR C4d- CAMR C4d+ TCMR Mixed Rejection BKV nephropathy IFTA Recurrence of initial nephropathy Other causes (septic shock, cardio-vascular diseases)

Supplementary Figure 3 A 100 * * ****

40 % of CD 8

0 M0 M12 M0 M12 M0 M12 NAÏVE EM TEMRA B **** **** **** **** **** **** 100 ** *** * *** 100 **** **** 100 **** ***

80 80 80 + + + 60 60 60

40 TBX21 40 40 % of % of GZM b 20 20 % of EOME S 20

0 0 0 EM EM EM EM EM EM NAÏVE NAÏVE NAÏVE NAÏVE NAÏVE NAÏVE TEMRA TEMRA TEMRA TEMRA TEMRA TEMRA

M0 M12 M0 M12 M0 M12

Supplementary Figure 4 % of CD69a+ 100 20 40 60 80 0

NAÏVE * M **

0 EM TEMRA

M12 NAÏVE * EM *** TEMRA Supplementary Figure 5 % of CD25+ within CD69+ 100 20 40 60 80 0 *

NAÏVE * M * * 0 EM TEMRA * M12

NAÏVE * ** EM TEMRA

% of CD107a+ within CD69+ 100 20 40 60 80 0

NAÏVE * M *

0 EM TEMRA

M12 NAÏVE * *

EM * TEMRA A Component 1 Component 2 IL23A SLAMF1 AQP3 IL15RA ABHD2 CDKN1A EPHA1 ST6GALNAC2 RPS2 PCNXL2 TGFBR3 RDX EEF1B2 C12orf23 LRRN3 GSTZ1 C17orf76−AS1 TRADD RPS5 SERPINB6 TPT1 GSTK1 EIF3E MPZL3 C12orf75 DUSP4 RSL1D1 RORA RGS9 RNF19A CD63 CD82 SPON2 LMNA FKBP5 WDR86−AS1 APEX1 ITPR3 PIK3AP1 ISG20 PRF1 PRKACB SLA2 EPHA4 PABPC1 ATXN1 ZNF101 MZT1 TACC3 MRI1 MYO6 TTC39C HAPLN3 COMMD1 FCRL6 PIK3R1 LITAF ZFP36L2 C14orf64 SURF4 GNPTAB DEGS1 LYN GAPDH KIR3DL2 FBXL8 BHLHE40 PRDX6 KIR3DL1 LUZP1 RPL13 BCDIN3D MCTP2 ATF6 SERINC5 MRPL1 CPT1A PAK1 HPCAL4 RPS6KA3 APMAP IL18R1 PLEK GZMK LEF1 TNFSF10 TNK1 MAP3K1 IMPDH2 ALOX5AP TGFBR1 LPIN2 SELL NCOA7 TARP OSBPL3 TCF7 AMICA1 FEZ1 ATP2B1 FAM60A PPIB FGFBP2 CITED2 PRR5L ANXA2 PLEKHB1 CANX NKG7 CTSA ENPP4 TPRG1 NGFRAP1 PERP RCAN3 GPR183 NPM1 MSC PTPN4 ITGA4 TMCC3 LOC100130476 TTC38 SNRNP200 MTSS1 FTH1 GIMAP5 ANKRD39 FAM65B FYN PDGFD PLAC8 SPINT2 ERN1 ZNF683 NCAPH NCALD CXCR6 CD244 EZR EFHD2 TRAF5 CD300A GPR15 S1PR5 PCED1B−AS1 MYC FYCO1 CCR7 FAM129A CD28 ATP10A GZMB CYB561 EEF1G ITPRIPL1 NELL2 CCR2 CD55 CCND2 TMEM66 CLU TBC1D4 ANXA1 DGKA PHACTR2 FAM49A PHLDA1 LEPROTL1 MBOAT1 ZEB2 CCR5 GZMH HNRPLL FCGR3A CCR4 USP28 EMB GPR56 EML4 MAL KIAA0319L CPNE2 PTPRC CX3CR1 RGS2 FGR GPR171 KLRD1 GALM PRSS23 LGALS3 ITGAM S100A11 LDHB CLDND1 GNLY CCR6 PASK FAS

−0.2 −0.1 0.0 0.1 0.2 −0.2 −0.1 0.0 0.1

Supplementary Figure 6A B NAÏVE EM TEMRA NAÏVE EM TEMRA GO Proliferation GO Signaling PTPRC FYN IL15RA RGS2 FYN FAM129A ISG20 PLEK CLU ERN1 S100A11 ITPR3 ZFP36L2 GPR171 CITED2 CXCR6 CCND2 PIK3R1 ATXN1 SLAMF1 PTPRC CDKN1A EPHA4 LEF1 TNFSF10 MYC RPS6KA3 IL23A NPM1 CD28 TRADD IMPDH2 CCR5 TACC3 CITED2 USP28 ITGA4 ZEB2 FAS ITGAM DUSP4 TGFBR1 CCND2 LYN CCR2 TGFBR3 TRAF5 GPR15 NPM1 MAP3K1 ANXA1 CDKN1A FTH1 KLRD1 GO Activation TGFBR3 FYN LITAF PLEK SLA2 PTPRC CX3CR1 CD63 GPR56 SLA2 ZEB2 RORA ITGAM CX3CR1 TGFBR1 ITGAM PDGFD FGR MTSS1 RGS9 LYN LYN ITGA4 LEF1 ATP2B1 MYC PIK3R1 LRRN3 CLU SERINC5 ITPR3 DGKA IMPDH2 CD28 LEF1 NGFRAP1 DGKA TBC1D4 IL23A TNK1 GIMAP5 EPHA1 GPR183 CD28 TCF7 GPR183 PRKACB SLAMF1 PAK1 CDKN1A ANXA1

ANXA1 GO Migration PAK1 FYN RGS2 GO Apoptosis APEX1 EPHA4 MAL PIK3R1 CD28 ATXN1 CD63 GIMAP5 PLEK LEF1 ITPR3 MYC TNFSF10 NGFRAP1 RPS6KA3 TPT1 TRADD PTPRC CCR5 TNFSF10 PERP FAS EMB CITED2 ERN1 ITGA4 TRADD PTPRC CLU EZR PERP SELL CITED2 CX3CR1 PHLDA1 GPR56 MAP3K1 ZEB2 CDKN1A CD244 FEZ1 USP28 LITAF CX3CR1 MYO6 TGFBR1 MTSS1 GZMB SPON2 NPM1 ITGAM ANXA1 TGFBR1

GO Immune Response RGS9 ITGA4 ABHD2 FAS LYN CCR5 CPT1A CLU CD300A PIK3R1 BHLHE40 RORA SLA2 TGFBR3 SELL LEF1 FYN SPINT2 PLEK HAPLN3 PTPRC AMICA1 IL18R1 NELL2 TNFSF10 CCR4 RPS6KA3 CCR2 SLA2 CCR6 CD244 MPZL3 KLRD1 TRAF5 TGFBR3 CCR7 CX3CR1 MAP3K1 FCGR3A PRKACB ITGAM PAK1 KIR3DL2 ANXA1 TGFBR1 SPON2 KIR3DL1 LYN FTH1 PAK1 ANXA1 AQP3 CD28 CD55 LEF1 IL23A IMPDH2 TACC3 TCF7 GIMAP5 AMICA1 PRF1 GPR183 CCR7 SLAMF1 CCR4 CCR2 CCR6 CDKN1A DUSP4 Supplementary Figure 6B A TEMRA CD16p B CD45RA CCR7 CD16 CD28 CD27 EM 18 21 17 CM 7 22 12 8 14 11 9 10 5 15 CD127 CD45RO CD56 CD25 CD95 2 21 23 6 20 16 13 4 3 CM EM 1 19

NAÏVE TEMRA CD16n CD62L CXCR3 CCR4 CCR5 CCR6

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 Rphenograph HLA-DR GZMB CD57 CD11a CD69

PD1 LAG3 CTLA4 BTLA

20 EM 23 TEMRA 21 EM 03 CM 02 NAÏVE 06 12 07 CM 11 14 EM 08 CM 15 19 13 NAÏVE 04 01 05 09 16 17 TEMRA 22 10 18 PD1 BTLA CD28 CD27 CD25 CD57 CD16 CD56 CD95 CD69 LAG3 CCR4 CCR6 CCR5 CCR7 GZMb CD11a CD62L CD127 CTLA4 CXCR3 HLA-DR CD45RA CD45RO

Supplementary Figure 7 A B NAÏVE EM TEMRA NAÏVE EM TEMRA

0.00 0.00 0.00 0.29 1.29 0.66

aCD16 aCD16 0.00 1.53 4.55 0.33 3.39 6.63

IL-15 0.00 0.00 0.00 IL-15 0.20 1.18 0.46

aCD16 aCD16 0.10 8.47 32.8 0.47 5.91 27.8 IL-15 IL-15 TNFa CD107a IFNg FSC

C D 5 60 * NAÏVE EM TEMRA 0.22 0.32 0.18 Raji 4 * + + 40 3

Raji 0.34 0.63 2.71 % of CD107 a

% of CD107 a 20 + RTX 1

0 CD107a RTX - + - + - + CD3 NAÏVE EM TEMRA aCD4 control aCD1 6 aHLA Class I E 30 **

% of CD107 a 10

0 control aCD1 6 aHLA Class I

Supplementary Figure 8 Target Fluorochrome Clone Name Provider CD3 VioGreen BW264/56 Miltenyi CD8 VioGreen BW135/80 Miltenyi CD27 VioBright FITC M-T271 Miltenyi CD45RA APC-Vio770 T6D11 Miltenyi CCR7 PE-Vio770 REA108 Miltenyi TNFa FITC cA2 Miltenyi PERF VioBlue Delta G9 Miltenyi T-Bet PE REA102 Miltenyi CD8 APC RPA-T8 BD Biosciences CD16 PE-Cy7 3G8 BD Biosciences CD3 PE-CF594 UCHT1 BD Biosciences CD3 FITC UCHT1 BD Biosciences CD25 FITC M-A251 BD Biosciences CD28 BV711 CD28.2 BD Biosciences CD57 BV605 NK-1 BD Biosciences CD69 PerCp-Cy5.5 FN50 BD Biosciences CD107a PE H4A3 BD Biosciences CCR7 BV421 150503 BD Biosciences CCR7 V450 150503 BD Biosciences IFNg V450 B27 BD Biosciences GZMb Alexa Fluor 700 GB11 BD Biosciences EOMES eFluor660 WD1928 Thermofisher

Supplementary Table 1. Listing of monoclonal antibodies Main GO ENRICH GO CATEGORY FDR Category MENT Proliferation GO:0008283_cell_proliferation 2.09 0.004091 GO:0008284_positive_regulation_of_cell_pr 2.45 0.048677 oliferation GO:0032943_mononuclear_cell_proliferatio 3.95 0.020492 n GO:0046651_lymphocyte_proliferation 3.95 0.020492 GO:0070661_leukocyte_proliferation 3.95 0.020492 Activation GO:0001775_cell_activation 3.29 0.000000 GO:0002694_regulation_of_leukocyte_activ 2.79 0.027771 ation GO:0042110_T_cell_activation 2.84 0.014989 GO:0045321_leukocyte_activation 3.16 0.000182 GO:0046649_lymphocyte_activation 3.10 0.000679 GO:0050865_regulation_of_cell_activation 3.03 0.014891 GO:0050867_positive_regulation_of_cell_ac 3.22 0.020415 tivation GO:0051251_positive_regulation_of_lymph 3.01 0.039481 ocyte_activation Apoptosis GO:0006916_anti-apoptosis 3.21 0.009962 GO:0006917_induction_of_apoptosis 2.57 0.011783 GO:0008633_activation_of_pro- 7.66 0.008149 apoptotic_gene_products GO:0010942_positive_regulation_of_cell_de 2.16 0.025574 ath GO:0012502_induction_of_programmed_cel 2.57 0.011783 l_death GO:0043065_positive_regulation_of_apopto 2.18 0.024336 sis GO:0043066_negative_regulation_of_apopto 2.45 0.018339 sis

1 GO:0043068_positive_regulation_of_progra 2.18 0.024336 mmed_cell_death GO:0043069_negative_regulation_of_progra 2.43 0.020345 mmed_cell_death GO:0060548_negative_regulation_of_cell_d 2.39 0.021040 eath Immune GO:0002250_adaptive_immune_response 3.90 0.034490 Response GO:0002252_immune_effector_process 3.15 0.011729 GO:0002253_activation_of_immune_respon 2.74 0.048390 se GO:0002376_immune_system_process 2.94 0.000000 GO:0002460_adaptive_immune_response_b ased_on_somatic_recombination_of_immune 3.96 0.033440 _receptors_built_from_immunoglobulin_sup erfamily_domains GO:0002520_immune_system_development 2.49 0.021085 GO:0002682_regulation_of_immune_system 2.75 0.000178 _process GO:0002684_positive_regulation_of_immun 2.49 0.017418 e_system_process GO:0006955_immune_response 3.48 0.000000 GO:0006959_humoral_immune_response 10.73 0.000065 GO:0050776_regulation_of_immune_respon 3.16 0.000654 se GO:0050778_positive_regulation_of_immun 2.60 0.039376 e_response Signaling GO:0001932_regulation_of_protein_phosph 3.45 0.015776 orylation GO:0007166_cell_surface_receptor_linked_s 2.44 0.000000 ignaling_pathway GO:0007186_G- protein_coupled_receptor_protein_signaling_ 5.07 0.000000 pathway

2 GO:0007243_intracellular_protein_kinase_c 2.19 0.017654 ascade GO:0023014_signal_transmission_via_phosp 2.19 0.017654 horylation_event GO:0033674_positive_regulation_of_kinase 3.37 0.007183 _activity GO:0042325_regulation_of_phosphorylation 2.33 0.016812 GO:0043085_positive_regulation_of_catalyti 2.55 0.002600 c_activity GO:0045859_regulation_of_protein_kinase_ 2.33 0.048429 activity GO:0045860_positive_regulation_of_protein 3.50 0.006353 _kinase_activity GO:0046777_protein_autophosphorylation 5.09 0.007014 GO:0051336_regulation_of_hydrolase_activi 2.45 0.029788 ty GO:0051338_regulation_of_transferase_acti 2.32 0.033336 vity GO:0051345_positive_regulation_of_hydrol 3.36 0.017664 ase_activity GO:0051347_positive_regulation_of_transfe 3.53 0.002594 rase_activity Migration GO:0006928_cellular_component_movemen 3.47 0.000000 t GO:0006935_chemotaxis 3.94 0.000067 GO:0007155_cell_adhesion 3.92 0.000000 GO:0010646_regulation_of_cell_communica 2.14 0.014956 tion GO:0016337_cell-cell_adhesion 4.29 0.015907 GO:0016477_cell_migration 3.75 0.000000 GO:0022610_biological_adhesion 3.92 0.000000 GO:0040011_locomotion 3.72 0.000000 GO:0048870_cell_motility 3.59 0.000037

3 GO:0050900_leukocyte_migration 4.60 0.002540 GO:0051270_regulation_of_cellular_compo 3.69 0.017639 nent_movement GO:0051674_localization_of_cell 3.59 0.000037 Other GO:0006816_calcium_ion_transport 4.77 0.017221 GO:0006873_cellular_ion_homeostasis 4.52 0.000040 GO:0006874_cellular_calcium_ion_homeost 7.03 0.000000 asis GO:0006875_cellular_metal_ion_homeostasi 6.86 0.000000 s GO:0007204_elevation_of_cytosolic_calciu 10.16 0.000000 m_ion_concentration GO:0030003_cellular_cation_homeostasis 5.54 0.000000 GO:0050801_ion_homeostasis 4.26 0.000150 GO:0051480_cytosolic_calcium_ion_homeo 9.19 0.000000 stasis GO:0055065_metal_ion_homeostasis 6.77 0.000000 GO:0055074_calcium_ion_homeostasis 6.92 0.000000 GO:0055080_cation_homeostasis 5.00 0.000154 GO:0072503_cellular_divalent_inorganic_ca 6.60 0.000042 tion_homeostasis GO:0072507_divalent_inorganic_cation_ho 6.50 0.000071 meostasis GO:0006952_defense_response 2.84 0.000143 GO:0006968_cellular_defense_response 9.10 0.000174 GO:0009605_response_to_external_stimulus 2.76 0.000186 GO:0009615_response_to_virus 3.35 0.011942 GO:0019724_B_cell_mediated_immunity 6.70 0.012011 GO:0019725_cellular_homeostasis 3.72 0.000063 GO:0030154_cell_differentiation 2.03 0.000852 GO:0048583_regulation_of_response_to_sti 2.84 0.000000 mulus

4 GO:0048584_positive_regulation_of_respon 2.56 0.018216 se_to_stimulus GO:0051130_positive_regulation_of_cellular 4.18 0.001232 _component_organization GO:0051174_regulation_of_phosphorus_met 2.37 0.011764 abolic_process GO:0080134_regulation_of_response_to_str 2.29 0.035572 ess

Supplementary Table 2. Gene ontology (GO) categories of CD8 subsets gene signature identified by sPLSDA