PD-1/TIGIT CD4⁺ T-Cell Signature for Risk Stratification of Post-Weaning Mortality in Venoarterial Extracorporeal Membrane Oxygenation Patients
Authors: Zengtao Wang, Xing Hao, Junyan Han, Chenglong Li, Liangshan Wang, Jing Chen, Jie Chen, Yu Cao, Yu Jiang, Haoyu Ji, Jiarui Li, Ju Zhang, Hong Wang, Mingxi Hua, Hui Zeng, Xiaotong Hou, Chen Chen
Published: 2026

ABSTRACT
BackgroundMortality after successful weaning from venoarterial extracorporeal membrane oxygenation (VA-ECMO) remains common, yet few immune biomarkers enable post-weaning risk stratification. We examined whether circulating CD4⁺ T-cell subsets defined by programmed cell death protein 1 (PD-1) and T cell immunoreceptor with immunoglobulin and immunoreceptor tyrosine-based inhibitory motif domains (TIGIT) were associated with off-ECMO mortality in adults supported with VA-ECMO.MethodsWe conducted flow cytometric profiling of circulating T cells in 125 adult VA-ECMO patients, classified as survivors (n = 70), on-ECMO mortality (n = 33), and off-ECMO mortality after successful weaning (n = 22). We first quantified CD3⁺, CD4⁺, and CD8⁺ T-cell counts using BD Trucount tubes, and assessed conventional CD4⁺ and CD8⁺ differentiation subsets by flow cytometry. We then quantified PD-1/TIGIT-defined CD4⁺ T cell subsets and assessed their ability to discriminate off-ECMO mortality among successfully weaned patients using receiver operating characteristic (ROC) analyses with bootstrap resampling. A PD-1/TIGIT-based risk classification was constructed using ROC-derived cut-offs.ResultsAlthough CD3⁺/CD4⁺/CD8⁺ T-cell counts and conventional CD4⁺ differentiation subsets did not differ across outcome groups, the off-ECMO mortality group showed a higher frequency of CD4⁺PD-1−TIGIT⁺ T cells [8.36% (5.65–11.78) vs 5.37% (3.90–7.15), p = 0.005] and a lower frequency of CD4⁺PD-1⁺TIGIT− T cells [10.90% (7.92–14.25) vs 14.95% (10.78–21.85), p = 0.017] compared with survivors. On day 1 of ECMO support, these subsets discriminated off-ECMO mortality with areas under the curve (AUCs) of 0.729 and 0.702, respectively. Based on ROC-derived cut-offs, off-ECMO mortality rates were 10.53%, 16.22%, and 70.59% in the low-, intermediate-, and high-risk groups.ConclusionA circulating CD4⁺ T cell signature defined by PD-1 and TIGIT expression associates with off-ECMO mortality and may aid risk stratification after VA-ECMO weaning.

Introduction

Venoarterial extracorporeal membrane oxygenation (VA-ECMO) provides temporary mechanical circulatory support for critically ill patients with refractory cardiac failure.1,2 Despite substantial advances in ECMO technology and clinical management, patients supported with ECMO continue to experience high rates of complications and mortality, which are associated with the severity of underlying illness, prolonged tissue hypoperfusion, coagulation disturbances, infection, and ECMO-related inflammatory and vascular injury.1,3–5 According to the 2024 Extracorporeal Life Support Organization (ELSO) registry, overall survival among adult cardiac patients receiving VA-ECMO is approximately 48%.6 Multiple organ failure represents the leading cause of death in ECMO patients and has been closely linked to ECMO-associated immune dysregulation.7,8 Multiple prognostic scoring systems, including the Survival After VA-ECMO (SAVE) score, the prEdictioN of Cardiogenic shock OUtcome foR Acute myocardial infarction patients salvaGed by VA-ECMO (ENCOURAGE) score, and the pRedicting mortality in patients undergoing veno-arterial Extracorporeal MEMBrane oxygenation after coronary artEry bypass gRafting (REMEMBER) score, have been developed to predict survival in VA-ECMO patients using pre-ECMO clinical and hemodynamic variables.9–11

Meanwhile, ECMO induces profound immune perturbations, highlighting the potential value of immune biomarkers as complementary prognostic tools. ECMO disrupts host immune homeostasis through multiple mechanisms, including direct blood contact with the artificial circuit, endothelial injury, and ischemia–reperfusion damage.12,13 These processes can affect several interconnected pathways, such as coagulation, complement activation, innate immune activation, and adaptive immune responses.14 Among these pathways, innate immune activation can drive rapid inflammatory responses; therefore, prior ECMO research has largely focused on innate immunity. These studies have examined pro-inflammatory cytokine release, shifts in neutrophil and monocyte populations, and changes in immune cell activation and adhesion molecule expression.15–18 Several of these innate immune alterations have been associated with thrombocytopenia and increased mortality in ECMO-supported patients.19,20

However, accumulating evidence indicates that sustained systemic inflammation also affects the adaptive immune system, particularly T cells, which are highly sensitive to ischemia–reperfusion injury and cytokine imbalance.21–24 In ECMO-supported patients, persistent lymphopenia, increased neutrophil-to-lymphocyte ratios, and reduced regulatory T cell counts associate with higher mortality.16,25,26 Together with reports of impaired T cell proliferation and increased apoptosis,27 these findings indicate significant T cell dysfunction during ECMO support.

One plausible mechanism underlying sustained T cell dysfunction is altered expression of inhibitory molecules, including programmed cell death protein 1 (PD-1), T cell immunoreceptor with immunoglobulin and immunoreceptor tyrosine-based inhibitory motif domains (TIGIT), and T-cell immunoglobulin and mucin domain-containing protein 3 (TIM-3). These molecules play essential roles in restraining excessive effector responses and preventing immune-mediated tissue damage.28 In chronic infections and cancer, persistent upregulation of inhibitory receptors characterizes T cell exhaustion and correlates with progressive loss of effector function.29 In patients undergoing cardiopulmonary bypass (CPB), a procedure similar to ECMO, altered PD-1 and TIM-3 expression on T cells has been linked to lung injury.30,31 Thus, the expression of inhibitory receptors on T cells in ECMO-supported patients also merits systematic investigation.

In this study, we performed flow cytometric analysis of circulating T cells in adult patients receiving VA-ECMO, focusing on dynamic changes in inhibitory receptor expression during ECMO support and after weaning, as well as their associations with clinical outcomes. Specifically, this study aimed to determine whether PD-1/TIGIT-defined CD4⁺ T cell subsets could provide complementary immune biomarker information for post-weaning risk stratification beyond conventional clinical assessment.

Materials and Methods

Study Population and Design

This observational study received approval from the Ethics Committee of Beijing Anzhen Hospital, Capital Medical University, Beijing, China (No. 2019027X), and was registered at ClinicalTrials.gov (NCT04158479). A total of 169 patients who received VA-ECMO at Beijing Anzhen Hospital were screened between May 2018 and October 2024. We excluded patients who were younger than 18 years (n = 1), lacked key clinical or laboratory data (n = 17), underwent a second VA-ECMO run (n = 1), received ECMO support for less than 24 hours (n = 19), or had fulminant myocarditis (n = 6).32,33 After applying these criteria, 125 adult patients receiving VA-ECMO were included in the final analysis (Figure 1A). Written informed consent was obtained from all participants or substitute decision makers prior to sample collection, in accordance with local regulations and the Declaration of Helsinki. Figure 1Study flow diagram and clinical outcomes of the VA-ECMO cohort. (A) Flow diagram showing patient enrollment, exclusions, and outcome classification following venoarterial extracorporeal membrane oxygenation (VA-ECMO) support. (B) Sankey diagram illustrating screening completeness and clinical outcomes of the VA-ECMO cohort.The image A shows a flow diagram detailing patient enrollment and outcomes for VA-ECMO support. Initially, 169 patients were supported by VA-ECMO. Exclusions included age under 18 (n = 1), lack of necessary parameters (n = 17), second ECMO (n = 1), VA-ECMO support under 24 hours (n = 19) and fulminant myocarditis (n = 6), totaling 44 exclusions. This left 125 adult patients. The VA-ECMO weaning procedure resulted in 92 successful weanings and 33 failed weanings. Among the successful, there were 70 30-day survivors and 22 non-survivors. The image B shows a Sankey diagram illustrating screening completeness and clinical outcomes. From the overall population of 169, 74.0 percent had complete screening data. Of the 125 on VA-ECMO, 73.6 percent were successfully weaned. Failed weaning resulted in 26.4 percent on-ECMO mortality. Among the weaned, 17.6 percent were 30-day non-survivors and 56.0 percent were 30-day survivors.Flow and Sankey diagrams of VA-ECMO patient outcomes and exclusions.

Study flow diagram and clinical outcomes of the VA-ECMO cohort. (A) Flow diagram showing patient enrollment, exclusions, and outcome classification following venoarterial extracorporeal membrane oxygenation (VA-ECMO) support. (B) Sankey diagram illustrating screening completeness and clinical outcomes of the VA-ECMO cohort.

Weaning was considered unsuccessful if a patient died while receiving ECMO support or required ECMO recannulation within 2 days after decannulation. Based on weaning status, patients were initially categorized into a successfully weaned group and a failed weaning group (on-ECMO mortality). The successfully weaned group was then further divided into 30-day survivors (survivors) and 30-day non-survivors (off-ECMO mortality) (Figure 1A and B).

In addition, 24 healthy controls were enrolled. Their health status was confirmed by physicians before enrollment. Blood samples were collected after written informed consent had been obtained. We excluded individuals with human immunodeficiency virus (HIV) infection, viral hepatitis, systemic infection, connective tissue disease, cancer, or abnormal tumor marker levels, including alpha-fetoprotein (AFP), carcinoembryonic antigen (CEA), carbohydrate antigen 19–9 (CA-199), CA-153, and CA-125.34

ECMO Implantation and Management

The details regarding VA-ECMO initiation and management have been described previously.35 VA-ECMO was initiated using peripheral femoral cannulation via either a percutaneous or surgical approach. To prevent severe limb ischemia, a 6 Fr distal perfusion catheter was routinely inserted into the superficial femoral artery. Trained members of the ECMO team performed all procedures. ECMO blood flow was adjusted according to continuous clinical assessment, including signs of tissue hypoperfusion, lactate clearance, and normalization of mean arterial pressure. Intravenous unfractionated heparin was administered to maintain an activated clotting time of 180–210 s or an activated partial thromboplastin time of 1.5–2.0 times the normal value. Patients were closely monitored for ECMO-related complications. Once patients met institutional weaning criteria, we conducted a standardized, protocol-driven weaning trial.36

Sample Collection

Peripheral blood samples were collected at predefined time points during ECMO support, including day 1 (D1), D2, D3, D5, and D7, and after ECMO weaning on post-weaning days 1 (P1), P2, P3, P5, and P7. All samples were drawn into EDTA-containing vacuum tubes as part of routine clinical testing. Immediately after collection, peripheral blood mononuclear cells (PBMCs) were isolated by density-gradient centrifugation using Ficoll-Paque (Amersham Pharmacia Biotech, Sweden).

Immunology Markers and Flow Cytometry Analysis

All experiments and assays were performed using freshly isolated samples. Isolated PBMCs were incubated with directly conjugated fluorescent antibodies for 30 min at 4 °C and then washed before flow cytometric analysis.34 T cells were identified by sequential gating on forward scatter (FSC) and side scatter (SSC) characteristics, followed by selection of CD3⁺ events. The following antibodies were used: anti-human CD3–Alexa Fluor 700, CD4–APC-Fire 750, CD8–BV605, PD-1–APC, and CCR7–BV421 (BioLegend, San Diego, CA, USA); CD45RA–BV510 (BD Biosciences, San Diego, CA, USA); and TIGIT–PE-Cy7 (eBioscience, San Diego, CA, USA), along with the corresponding isotype controls. Data were acquired on a BD FACSCanto™ flow cytometer (BD Biosciences) and analyzed using FlowJo software (Tree Star, Ashland, OR, USA).

Absolute leukocyte counts were determined using BD Trucount Tubes (BD Biosciences) in combination with a CD45/CD3/CD4/CD8 antibody cocktail (BD Biosciences), following the manufacturer’s instructions. Absolute T cell counts (cells per microliter) were calculated by relating cellular events to bead events within each Trucount tube. For these measurements, data were acquired on a FACSCalibur™ flow cytometer (BD Biosciences) and analyzed using FlowJo software (Tree Star, Ashland, OR, USA).

Statistical Analysis

Supplementary Figure 1 outlines the overall analytical workflow. Among patients who were successfully weaned from VA-ECMO, we examined the associations between day 1 frequencies of CD4⁺PD-1−TIGIT⁺ and CD4⁺PD-1⁺TIGIT− T cells and off-ECMO mortality. Receiver operating characteristic (ROC) curves were generated for each subset, and the areas under the curve (AUCs) were calculated using SPSS to evaluate discriminative performance. Optimal cut-off values for CD4⁺PD-1−TIGIT⁺ and CD4⁺PD-1⁺TIGIT− frequencies were identified by maximizing the Youden index.

To test the robustness of these associations, interval validation was performed using bootstrap resampling with 1000 iterations among successfully weaned patients. In each iteration, patients were resampled with replacement, and ROC curves and AUCs were recalculated for each marker. Bootstrap distributions were used to derive 95% confidence intervals for AUCs and to assess the stability of differences in marker frequencies between clinical outcome groups.

A tripartite risk stratification was constructed based on two ROC-derived cut-offs. Patients with CD4⁺PD-1−TIGIT⁺ frequencies < 6.71% and CD4⁺PD-1⁺TIGIT− frequencies > 11.65% of total CD4⁺ T cells were categorized as low risk, and those with CD4⁺PD-1−TIGIT⁺ > 6.71% and CD4⁺PD-1⁺TIGIT− < 11.65% were categorized as high risk. All remaining patients—those with either both values above or both values below the cut-offs—were assigned to the intermediate-risk group. Off-ECMO mortality rates were compared across risk groups using Fisher’s exact test.

To further explore whether PD-1/TIGIT-defined variables were associated with off-ECMO mortality after successful weaning, additional univariable and multivariable logistic regression analyses were performed among successfully weaned patients, with off-ECMO mortality as the outcome. Candidate covariates were selected based on previously reported predictors of post-weaning outcomes and variables associated with off-ECMO mortality in the current cohort.37 Multivariable models were adjusted for age, CRRT, IABP, diabetes, LVEF, and lactate levels at 4 and 24 hours after ECMO initiation. Each PD-1/TIGIT-defined variable was entered separately into the multivariable model. Given the limited number of off-ECMO mortality events, these analyses were considered exploratory.

All statistical analyses were performed using R, SPSS (IBM Corporation, New York, NY, USA), and GraphPad Prism 9 (GraphPad Software, La Jolla, CA, USA). Continuous variables are presented as medians with interquartile ranges, and categorical variables are presented as counts with percentages. For comparisons between two independent groups, we analyzed continuous variables using unpaired two-tailed Student’s t-tests for normally distributed data or the Mann–Whitney U-test for non-normally distributed data. For comparisons among three groups, we used one-way ANOVA with Tukey’s post hoc test for normally distributed variables and the Kruskal–Wallis test with Dunn’s post hoc test for non-normally distributed variables. Categorical variables were compared using the χ2-test or Fisher’s exact test, as appropriate. p ≤ 0.05 represents statistically significant.

Results

Clinical Characteristics of Cohorts

In this study, we screened 169 patients who received VA-ECMO support and included 125 adult patients in the final analysis (Figure 1A and B). Of these patients, 70 (56.0%) were successfully weaned from ECMO and survived for 30 days (survivors), 33 (26.4%) patients failed to wean (on-ECMO mortality), and 22 (17.6%) died within 30 days after successful weaning (off-ECMO mortality). Table 1 summarizes baseline clinical characteristics and clinical outcomes.Table 1The Demographic Information and Characteristics of Patients Supported with ECMO (Survivors, off-ECMO Mortality, and on-ECMO Mortality)VariablesTotal (n = 125)Survivors (n = 70)Off-ECMO Mortality (n = 22)On-ECMO Mortality (n = 33)P valueAge, median (IQR), yr60.0 (52.0, 67.0)59 (51.0, 68.0)63.0 (55.0, 67.0)61.0 (52.0, 65.0)0.551BMI, median (IQR)24.8 (22.6, 26.9)24.6 (22.5, 26.7)26.1 (23.3, 27.3)25 (21.8, 27.0)0.292Female, n (%)31 (24.8)14 (20.0)5 (22.7)12 (36.4)0.194Major diagnosis on admissionCAD, n (%)80 (64.0)47 (67.1)15 (68.2)18 (54.5)0.417MI, n (%)33 (26.4)22 (31.4)6 (27.3)5 (15.2)0.216AD, n (%)11 (8.8)6 (8.6)1 (4.5)4 (12.1)0.621Valvular heart disease, n (%)37 (29.6)18 (25.7)7 (31.8)12 (36.4)0.526Echocardiography on admissionLVEF, median (IQR), %52.5 (32.8, 62.3)52.0 (33.0, 62.0)43.0 (21.8, 60.0)58.0 (38.5, 65.5)0.164LVEDD, median (IQR), mm47.0 (42.0, 56.0)49.0 (42.8, 56.0)45.5 (38.8, 55.3)48.0 (42.5, 56.5)0.632NYHA classification on admission, n (%)0.139Class I22 (18.5)15 (23.1)3 (14.3)4 (12.1)Class II23 (19.3)17 (26.2)1 (4.8)5 (15.2)Class III55 (46.2)23 (35.4)13 (61.9)19 (57.6)Class IV19 (16.0)10 (15.4)4 (19.0)5 (15.2)Medical historyCardiac surgery, n (%)11 (8.8)8 (11.4)2 (9.1)1 (3.0)0.373PCI, n (%)32 (25.6)15 (21.4)8 (36.4)9 (27.3)0.363MI, n (%)20 (16.0)11 (15.7)5 (22.7)4 (12.1)0.573Hypertension, n (%)63 (50.4)34 (48.6)14 (63.6)15 (45.5)0.376Hyperlipidemia, n (%)44 (35.2)26 (37.1)7 (31.8)11 (33.3)0.871Diabetes, n (%)34 (27.4)19 (27.1)9 (42.9)6 (18.2)0.140Cardiac surgery0.615CABG, n (%)42 (33.6)21 (30.0)10 (45.5)11 (33.3)CABG with Valve Surgery, n (%)13 (10.4)7 (10.0)2 (9.1)4 (12.1)Valve Surgery, n (%)21 (16.8)13 (18.6)1 (4.5)7 (21.2)Aortic Surgery, n (%)11 (8.8)6 (8.6)1 (4.5)4 (12.1)Heart transplantation, n (%)3 (2.4)3 (4.3)0 (0)0 (0)Others, n (%)35 (28.0)20 (28.6)8 (36.4)7 (21.2)Cardiac surgery on CPB, n (%)68 (54.8)37 (53.6)10 (45.5)21 (63.6)0.396CA prior to ECMO, n (%)25 (20.0)15 (21.4)3 (13.6)7 (21.2)0.713SOFA score before ECMO, median (IQR)8.0 (6.0, 10.0)8.0 (6.0, 10.0)8.0 (6.0, 11.0)9.0 (7.0, 11.0)0.327ECMO flow, median (IQR), L/min4 hours after ECMO3.3 (2.9, 3.6)3.1 (2.7, 3.5)3.5 (3.0, 3.9)3.4 (3.1, 3.7)0.03124 hours after ECMO3.2 (2.9, 3.7)3.1 (2.8, 3.6)3.4 (2.9, 3.8)3.4 (3.1, 3.8)0.048Lactate, median (IQR), mmol/L4 hours after ECMO7.9 (3.7, 14.3)6.3 (3.0, 12.2)7.3 (4.3, 16.3)11.2 (7.7, 19.0)0.00624 hours after ECMO3.3 (2.0, 6.0)2.3 (1.5, 4.2)3.7 (3.1, 7.6)7.5 (3.2, 12.8)<0.001Combined CRRT, n (%)45 (36.0)17 (24.3)9 (40.9)19 (57.6)0.004Combined IABP, n (%)64 (51.2)29 (41.4)15 (68.2)20 (60.6)0.041OutcomesBleeding, n (%)20 (16.1)8 (11.4)6 (28.6)6 (18.2)0.161Limb Complications, n (%)38 (30.4)17 (24.3)7 (31.8)14 (42.4)0.173Pulmonary infection, n (%)48 (38.4)25 (35.7)5 (22.7)18 (54.5)0.047Duration of MV, median (IQR), hours174.0 (78.5, 267.0)180.4 (102.6, 291.2)225.3 (179.5, 402.6)79.8 (52.9, 172.5)<0.001Duration of ECMO, median (IQR), hours92.8 (51.9, 159.4)91.8 (60.0, 144.8)173.5 (111.3, 256.1)60.1 (38.7, 110.2)<0.001Hospital LOS, median (IQR), days17 (12, 25)19 (14, 30)18 (14, 23)11 (7, 17)<0.001ICU LOS, median (IQR), days9 (5, 14)11 (7, 15)10 (6, 17)4 (2, 8)<0.001Abbreviations: AD, Aortic dissection; BMI, Body mass index; CA, cardiac arrest; CABG, coronary artery bypass grafting; CPB, cardiopulmonary bypass; CRRT, continuous renal replacement therapy; ECMO, extracorporeal membrane oxygenation; IABP, intra-aortic balloon pump; LOS, length of stay; LVEDD, left ventricular end diastolic diameter; LVEF, left ventricular ejection fraction; MI, myocardial infarct; MV, mechanical ventilation; NYHA classification, New York Heart Association classification; PCI, percutaneous coronary intervention; SOFA, the Sequential Organ Failure Assessment.

Table 1The Demographic Information and Characteristics of Patients Supported with ECMO (Survivors, off-ECMO Mortality, and on-ECMO Mortality)VariablesTotal (n = 125)Survivors (n = 70)Off-ECMO Mortality (n = 22)On-ECMO Mortality (n = 33)P valueAge, median (IQR), yr60.0 (52.0, 67.0)59 (51.0, 68.0)63.0 (55.0, 67.0)61.0 (52.0, 65.0)0.551BMI, median (IQR)24.8 (22.6, 26.9)24.6 (22.5, 26.7)26.1 (23.3, 27.3)25 (21.8, 27.0)0.292Female, n (%)31 (24.8)14 (20.0)5 (22.7)12 (36.4)0.194Major diagnosis on admissionCAD, n (%)80 (64.0)47 (67.1)15 (68.2)18 (54.5)0.417MI, n (%)33 (26.4)22 (31.4)6 (27.3)5 (15.2)0.216AD, n (%)11 (8.8)6 (8.6)1 (4.5)4 (12.1)0.621Valvular heart disease, n (%)37 (29.6)18 (25.7)7 (31.8)12 (36.4)0.526Echocardiography on admissionLVEF, median (IQR), %52.5 (32.8, 62.3)52.0 (33.0, 62.0)43.0 (21.8, 60.0)58.0 (38.5, 65.5)0.164LVEDD, median (IQR), mm47.0 (42.0, 56.0)49.0 (42.8, 56.0)45.5 (38.8, 55.3)48.0 (42.5, 56.5)0.632NYHA classification on admission, n (%)0.139Class I22 (18.5)15 (23.1)3 (14.3)4 (12.1)Class II23 (19.3)17 (26.2)1 (4.8)5 (15.2)Class III55 (46.2)23 (35.4)13 (61.9)19 (57.6)Class IV19 (16.0)10 (15.4)4 (19.0)5 (15.2)Medical historyCardiac surgery, n (%)11 (8.8)8 (11.4)2 (9.1)1 (3.0)0.373PCI, n (%)32 (25.6)15 (21.4)8 (36.4)9 (27.3)0.363MI, n (%)20 (16.0)11 (15.7)5 (22.7)4 (12.1)0.573Hypertension, n (%)63 (50.4)34 (48.6)14 (63.6)15 (45.5)0.376Hyperlipidemia, n (%)44 (35.2)26 (37.1)7 (31.8)11 (33.3)0.871Diabetes, n (%)34 (27.4)19 (27.1)9 (42.9)6 (18.2)0.140Cardiac surgery0.615CABG, n (%)42 (33.6)21 (30.0)10 (45.5)11 (33.3)CABG with Valve Surgery, n (%)13 (10.4)7 (10.0)2 (9.1)4 (12.1)Valve Surgery, n (%)21 (16.8)13 (18.6)1 (4.5)7 (21.2)Aortic Surgery, n (%)11 (8.8)6 (8.6)1 (4.5)4 (12.1)Heart transplantation, n (%)3 (2.4)3 (4.3)0 (0)0 (0)Others, n (%)35 (28.0)20 (28.6)8 (36.4)7 (21.2)Cardiac surgery on CPB, n (%)68 (54.8)37 (53.6)10 (45.5)21 (63.6)0.396CA prior to ECMO, n (%)25 (20.0)15 (21.4)3 (13.6)7 (21.2)0.713SOFA score before ECMO, median (IQR)8.0 (6.0, 10.0)8.0 (6.0, 10.0)8.0 (6.0, 11.0)9.0 (7.0, 11.0)0.327ECMO flow, median (IQR), L/min4 hours after ECMO3.3 (2.9, 3.6)3.1 (2.7, 3.5)3.5 (3.0, 3.9)3.4 (3.1, 3.7)0.03124 hours after ECMO3.2 (2.9, 3.7)3.1 (2.8, 3.6)3.4 (2.9, 3.8)3.4 (3.1, 3.8)0.048Lactate, median (IQR), mmol/L4 hours after ECMO7.9 (3.7, 14.3)6.3 (3.0, 12.2)7.3 (4.3, 16.3)11.2 (7.7, 19.0)0.00624 hours after ECMO3.3 (2.0, 6.0)2.3 (1.5, 4.2)3.7 (3.1, 7.6)7.5 (3.2, 12.8)<0.001Combined CRRT, n (%)45 (36.0)17 (24.3)9 (40.9)19 (57.6)0.004Combined IABP, n (%)64 (51.2)29 (41.4)15 (68.2)20 (60.6)0.041OutcomesBleeding, n (%)20 (16.1)8 (11.4)6 (28.6)6 (18.2)0.161Limb Complications, n (%)38 (30.4)17 (24.3)7 (31.8)14 (42.4)0.173Pulmonary infection, n (%)48 (38.4)25 (35.7)5 (22.7)18 (54.5)0.047Duration of MV, median (IQR), hours174.0 (78.5, 267.0)180.4 (102.6, 291.2)225.3 (179.5, 402.6)79.8 (52.9, 172.5)<0.001Duration of ECMO, median (IQR), hours92.8 (51.9, 159.4)91.8 (60.0, 144.8)173.5 (111.3, 256.1)60.1 (38.7, 110.2)<0.001Hospital LOS, median (IQR), days17 (12, 25)19 (14, 30)18 (14, 23)11 (7, 17)<0.001ICU LOS, median (IQR), days9 (5, 14)11 (7, 15)10 (6, 17)4 (2, 8)<0.001Abbreviations: AD, Aortic dissection; BMI, Body mass index; CA, cardiac arrest; CABG, coronary artery bypass grafting; CPB, cardiopulmonary bypass; CRRT, continuous renal replacement therapy; ECMO, extracorporeal membrane oxygenation; IABP, intra-aortic balloon pump; LOS, length of stay; LVEDD, left ventricular end diastolic diameter; LVEF, left ventricular ejection fraction; MI, myocardial infarct; MV, mechanical ventilation; NYHA classification, New York Heart Association classification; PCI, percutaneous coronary intervention; SOFA, the Sequential Organ Failure Assessment.

The Demographic Information and Characteristics of Patients Supported with ECMO (Survivors, off-ECMO Mortality, and on-ECMO Mortality)

Abbreviations: AD, Aortic dissection; BMI, Body mass index; CA, cardiac arrest; CABG, coronary artery bypass grafting; CPB, cardiopulmonary bypass; CRRT, continuous renal replacement therapy; ECMO, extracorporeal membrane oxygenation; IABP, intra-aortic balloon pump; LOS, length of stay; LVEDD, left ventricular end diastolic diameter; LVEF, left ventricular ejection fraction; MI, myocardial infarct; MV, mechanical ventilation; NYHA classification, New York Heart Association classification; PCI, percutaneous coronary intervention; SOFA, the Sequential Organ Failure Assessment.

T Cell Counts and Compositions in ECMO-Supported Patients with Different Outcomes

To examine the effects of ECMO on T cells, we analyzed longitudinal changes in peripheral T cell counts using flow cytometry. Consistent with prior reports, patients receiving ECMO exhibited reduced counts of total T cells, CD4⁺ T cells, and CD8⁺ T cells throughout ECMO support (Supplementary Figure 2A), with lymphopenia persisting for up to 7 days after weaning (Supplementary Figure 2B). In parallel, ECMO-supported patients showed a higher CD4/CD8 ratio than healthy controls (HCs) both during ECMO support and after weaning (Supplementary Figure 2C).

Because death occurred either during ECMO support or after successful weaning, we stratified patients into survivors, on-ECMO mortality, and off-ECMO mortality groups. These groups demonstrated similar longitudinal patterns in T cell counts and CD4/CD8 ratios throughout ECMO support and within one week after weaning (Supplementary Table 1, Supplementary Figure 3A and B). We then assessed T cell compartments, including naïve T cells (TN, CCR7⁺CD45RA⁺), central memory T cells (TCM, CCR7⁺CD45RA−), effector memory T cells (TEM, CCR7−CD45RA−), and terminally differentiated effector cells (TEMRA, CCR7−CD45RA⁺), within CD4⁺ and CD8⁺ T cells (Supplementary Figure 4A). Although ECMO-supported patients exhibited lower CD4⁺ TCM frequencies and higher CD8⁺ TN frequencies overall (Supplementary Figure 4B and C), the three outcome groups showed comparable distributions of these T cell subpopulations (Supplementary Figure 4D).

Alterations of TIGIT+ and PD-1+ T Cell Subpopulations in ECMO-Supported Patients

Next, we profiled co-inhibitory, co-stimulatory, and activation markers on T cells in ECMO-supported patients. Based on our pilot data identifying PD-1 and TIGIT as the most prominently altered markers, we quantified PD-1 and TIGIT expression on T cells by flow cytometry. We further subdivided CD4⁺ T cells into four subsets according to PD-1 and TIGIT expression: CD4⁺PD-1−TIGIT−, CD4⁺PD-1−TIGIT⁺, CD4⁺PD-1⁺TIGIT−, and CD4⁺PD-1⁺TIGIT⁺ cells (Figure 2A). Figure 2PD-1/TIGIT-defined T-cell subsets and differentiation compositions in VA-ECMO-supported patients. (A) Representative flow-cytometry plots showing PD-1 and TIGIT expression in circulating CD4⁺ and CD8⁺ T cells from healthy controls (HC) and VA-ECMO-supported patients (ECMO). (B) Frequencies of CD4⁺PD-1−TIGIT⁺ and CD4⁺PD-1⁺TIGIT− T cells among CD4⁺ T cells in HCs (n = 19, blue) and VA-ECMO patients (n = 125, yellow). (C) Differentiation composition (TN, blue; TCM, yellow; TEM, grey; TEMRA, red) within PD-1/TIGIT-defined CD4⁺ T-cell subsets (CD4⁺PD-1−TIGIT−, CD4⁺PD-1−TIGIT⁺, CD4⁺PD-1⁺TIGIT−, CD4⁺PD-1⁺TIGIT⁺) of HCs (n = 24). (D) Differentiation composition (TN, blue; TCM, yellow; TEM, grey; TEMRA, red) within PD-1/TIGIT-defined CD4⁺ T-cell subsets (CD4⁺PD-1−TIGIT−, CD4⁺PD-1−TIGIT⁺, CD4⁺PD-1⁺TIGIT−, CD4⁺PD-1⁺TIGIT⁺) of VA-ECMO patients (n = 36). Boxplots show medians with interquartile ranges (IQR). Age- and sex-matched HC subsets are used where applicable. P-values are indicated in the figure; two-group comparisons were performed using the Mann–Whitney U-test, and multi-group comparisons were performed using the Kruskal–Wallis test followed by Dunn’s multiple-comparisons correction.The study examines PD-1/TIGIT T-cell subsets in VA-ECMO patients using flow-cytometry. Image A displays density plots of PD-1 vs. TIGIT for HC and ECMO groups, divided into quadrants with numeric annotations. Image B shows boxplots comparing the frequencies of CD4⁺PD-1⁻TIGIT⁺ and CD4⁺PD-1⁺TIGIT⁻ cells among CD4⁻ T cells between HC and ECMO groups. Images C and D show the differentiation composition of four PD-1/TIGIT-defined CD4⁻ T-cell subsets in HC and ECMO groups, respectively. Each subset is divided into TN, TCM, TEM and TEMRA populations.A multi-panel flow-cytometry and boxplot figure comparing PD-1 and TIGIT T-cell subsets between groups.Abbreviations: VA-ECMO, venoarterial extracorporeal membrane oxygenation; HC, healthy control; TN, naïve T cells; TCM, central memory T cells; TEM, effector memory T cells; TEMRA, terminally differentiated effector memory T cells re-expressing CD45RA.

PD-1/TIGIT-defined T-cell subsets and differentiation compositions in VA-ECMO-supported patients. (A) Representative flow-cytometry plots showing PD-1 and TIGIT expression in circulating CD4⁺ and CD8⁺ T cells from healthy controls (HC) and VA-ECMO-supported patients (ECMO). (B) Frequencies of CD4⁺PD-1−TIGIT⁺ and CD4⁺PD-1⁺TIGIT− T cells among CD4⁺ T cells in HCs (n = 19, blue) and VA-ECMO patients (n = 125, yellow). (C) Differentiation composition (TN, blue; TCM, yellow; TEM, grey; TEMRA, red) within PD-1/TIGIT-defined CD4⁺ T-cell subsets (CD4⁺PD-1−TIGIT−, CD4⁺PD-1−TIGIT⁺, CD4⁺PD-1⁺TIGIT−, CD4⁺PD-1⁺TIGIT⁺) of HCs (n = 24). (D) Differentiation composition (TN, blue; TCM, yellow; TEM, grey; TEMRA, red) within PD-1/TIGIT-defined CD4⁺ T-cell subsets (CD4⁺PD-1−TIGIT−, CD4⁺PD-1−TIGIT⁺, CD4⁺PD-1⁺TIGIT−, CD4⁺PD-1⁺TIGIT⁺) of VA-ECMO patients (n = 36). Boxplots show medians with interquartile ranges (IQR). Age- and sex-matched HC subsets are used where applicable. P-values are indicated in the figure; two-group comparisons were performed using the Mann–Whitney U-test, and multi-group comparisons were performed using the Kruskal–Wallis test followed by Dunn’s multiple-comparisons correction.

Compared with healthy controls (HCs), ECMO-supported patients showed a reciprocal shift in the CD4⁺ compartment on day 1 of ECMO support, with reduced frequencies of CD4⁺PD-1−TIGIT⁺ cells and increased frequencies of CD4⁺PD-1⁺TIGIT− cells (Figure 2B), whereas the frequencies of CD4⁺PD-1−TIGIT− and CD4⁺PD-1⁺TIGIT⁺ cells were comparable between groups (Supplementary Figure 5A). This pattern was not observed in CD8⁺ T cells; although CD8⁺PD-1−TIGIT⁺ cells were reduced, other PD-1/TIGIT-defined CD8⁺ subsets showed similar frequencies between groups (Supplementary Figure 5B).

The differentiation profiles of CD4⁺PD-1−TIGIT⁺ and CD4⁺PD-1⁺TIGIT− subsets also differed between HCs and ECMO-supported patients. In HCs, CD4⁺PD-1−TIGIT⁺ cells were mainly TCM, whereas CD4⁺PD-1⁺TIGIT− cells consisted largely of TCM and TEM (Figure 2C). In ECMO-supported patients, CD4⁺PD-1−TIGIT⁺ cells comprised both TCM and TEM, while CD4⁺PD-1⁺TIGIT− cells were enriched for TEM (Figure 2D).

Discriminative Performance of CD4⁺PD-1−TIGIT⁺ and CD4⁺PD-1⁺TIGIT− T Cell Frequencies for off-ECMO Mortality

We next assessed whether alterations in PD-1/TIGIT-defined T cell subsets varied according to patient outcomes. On day 1 of ECMO support, survivors and non-survivors (including on-ECMO and off-ECMO mortality) showed similar frequencies of CD4⁺ and CD8⁺ PD-1−TIGIT⁺ and PD-1⁺TIGIT− subsets (Supplementary Figure 6A and B), indicating that these changes did not initially distinguish overall survival. In contrast, the off-ECMO mortality group displayed a higher frequency of CD4⁺PD-1−TIGIT⁺ T cells [8.36% (5.65–11.78) vs 5.37% (3.90–7.15), p = 0.005] and a lower frequency of CD4⁺PD-1⁺TIGIT− T cells [10.90% (7.92–14.25) vs 14.95% (10.78–21.85), p = 0.017] compared with survivors (Figure 3A). We observed no differences in CD4⁺PD-1−TIGIT− or CD4⁺PD-1⁺TIGIT⁺ subsets (Supplementary Figure 6C), nor in any CD8⁺ PD-1/TIGIT-defined subsets (Supplementary Figure 6D). Figure 3Dynamic changes and prognostic performance of PD-1/TIGIT-defined CD4⁺ T-cell subsets in VA-ECMO patients. (A) Frequencies of CD4⁺PD-1−TIGIT⁺ and CD4⁺PD-1⁺TIGIT− T cells among CD4⁺ T cells in VA-ECMO patients stratified as survivors (n = 70, red), on-ECMO mortality (n = 33, purple), and off-ECMO mortality after successful weaning (n = 22, yellow). (B and C) Dynamic trends of CD4⁺PD-1−TIGIT⁺ and CD4⁺PD-1⁺TIGIT− T cells (loess-smoothed curves with 95% confidence intervals) over days 1–3 of VA-ECMO support (D1-D3) (B) and post-weaning days 1–3 (P1-P3) (C) in patients stratified as survivors (n = 70, red), on-ECMO mortality (n = 33, purple), and off-ECMO mortality after successful weaning (n = 22, yellow). (D) Receiver operating characteristic (ROC) curves for discriminating off-ECMO mortality among successfully weaned patients (n = 92) using frequencies of CD4⁺PD-1−TIGIT⁺ and CD4⁺PD-1⁺TIGIT− cells, respectively. Solid lines indicate ROC curves from the original dataset, with the Youden index–derived optimal cut-offs annotated (6.71% for CD4⁺PD-1−TIGIT⁺ and 11.65% for CD4⁺PD-1⁺TIGIT−). Shaded bands represent bootstrap interval validation based on 1000 resamples. (E) PD-1/TIGIT-based risk classification of successfully weaned patients using ROC-derived cut-offs (6.71% for CD4⁺PD-1−TIGIT⁺ and 11.65% for CD4⁺PD-1⁺TIGIT−). Patients are categorized as low-risk (CD4⁺PD-1−TIGIT⁺ < 6.71% and CD4⁺PD-1⁺TIGIT− > 11.65%), high-risk (CD4⁺PD-1−TIGIT⁺ > 6.71% and CD4⁺PD-1⁺TIGIT− < 11.65%), or intermediate-risk (all remaining patients). Numbers denote off-ECMO deaths/total patients in each region. (F) Off-ECMO mortality rates across PD-1/TIGIT-based risk groups. Percentages above bars indicate mortality rates. P-values for group differences are shown. Boxplots show medians with interquartile ranges (IQR).Image A shows boxplots comparing the frequencies of CD4⁺PD-1−TIGIT⁺ and CD4⁺PD-1⁺TIGIT− T cells among survivors, patients with on-ECMO mortality, and patients with off-ECMO mortality. The off-ECMO mortality group shows higher CD4⁺PD-1−TIGIT⁺ frequencies and lower CD4⁺PD-1⁺TIGIT− frequencies than survivors. Images B and C show longitudinal trends of these two subsets during ECMO support and after weaning, respectively, with individual data points, smoothed curves, and confidence bands for each outcome group. Image D shows receiver operating characteristic curves for discriminating off-ECMO mortality, with AUC values of 0.729 for CD4⁺PD-1−TIGIT⁺ cells and 0.702 for CD4⁺PD-1⁺TIGIT− cells. Image E shows the PD-1/TIGIT-based risk classification using cut-offs of 6.71% for CD4⁺PD-1−TIGIT⁺ cells and 11.65% for CD4⁺PD-1⁺TIGIT− cells. Off-ECMO mortality cases are more frequent in the high-risk region. Image F shows off-ECMO mortality rates across the low-, intermediate-, and high-risk groups, increasing from 10.53% to 16.22% and 70.59%, respectively.A multi-panel figure showing dynamic changes and prognostic performance of PD-1/TIGIT-defined CD4⁺ T-cell subsets in VA-ECMO patients.

Dynamic changes and prognostic performance of PD-1/TIGIT-defined CD4⁺ T-cell subsets in VA-ECMO patients. (A) Frequencies of CD4⁺PD-1−TIGIT⁺ and CD4⁺PD-1⁺TIGIT− T cells among CD4⁺ T cells in VA-ECMO patients stratified as survivors (n = 70, red), on-ECMO mortality (n = 33, purple), and off-ECMO mortality after successful weaning (n = 22, yellow). (B and C) Dynamic trends of CD4⁺PD-1−TIGIT⁺ and CD4⁺PD-1⁺TIGIT− T cells (loess-smoothed curves with 95% confidence intervals) over days 1–3 of VA-ECMO support (D1-D3) (B) and post-weaning days 1–3 (P1-P3) (C) in patients stratified as survivors (n = 70, red), on-ECMO mortality (n = 33, purple), and off-ECMO mortality after successful weaning (n = 22, yellow). (D) Receiver operating characteristic (ROC) curves for discriminating off-ECMO mortality among successfully weaned patients (n = 92) using frequencies of CD4⁺PD-1−TIGIT⁺ and CD4⁺PD-1⁺TIGIT− cells, respectively. Solid lines indicate ROC curves from the original dataset, with the Youden index–derived optimal cut-offs annotated (6.71% for CD4⁺PD-1−TIGIT⁺ and 11.65% for CD4⁺PD-1⁺TIGIT−). Shaded bands represent bootstrap interval validation based on 1000 resamples. (E) PD-1/TIGIT-based risk classification of successfully weaned patients using ROC-derived cut-offs (6.71% for CD4⁺PD-1−TIGIT⁺ and 11.65% for CD4⁺PD-1⁺TIGIT−). Patients are categorized as low-risk (CD4⁺PD-1−TIGIT⁺ < 6.71% and CD4⁺PD-1⁺TIGIT− > 11.65%), high-risk (CD4⁺PD-1−TIGIT⁺ > 6.71% and CD4⁺PD-1⁺TIGIT− < 11.65%), or intermediate-risk (all remaining patients). Numbers denote off-ECMO deaths/total patients in each region. (F) Off-ECMO mortality rates across PD-1/TIGIT-based risk groups. Percentages above bars indicate mortality rates. P-values for group differences are shown. Boxplots show medians with interquartile ranges (IQR).

Because CD4⁺PD-1−TIGIT⁺ and CD4⁺PD-1⁺TIGIT− frequencies did not differ between survivors and on-ECMO mortality group (Figure 3A), these subsets showed limited value for identifying on-ECMO mortality. We therefore focused subsequent analyses on off-ECMO mortality among successfully weaned patients. Notably, elevated CD4⁺PD-1−TIGIT⁺ frequencies and reduced CD4⁺PD-1⁺TIGIT− frequencies persisted in the off-ECMO mortality group throughout ECMO support (D1–D3) and after weaning (P1–P3) (Figure 3B and C).

We then evaluated the ability of day 1 frequencies of these subsets to discriminate off-ECMO mortality using ROC analysis. CD4⁺PD-1−TIGIT⁺ cells yielded an AUC of 0.729 (95% CI, 0.608–0.849; p = 0.001), while CD4⁺PD-1⁺TIGIT− cells yielded an AUC of 0.702 (95% CI, 0.576–0.827; p = 0.004) (Figure 3D). An interval validation using bootstrap resampling (n = 1000) produced identical AUCs of 0.729 (95% CI, 0.595–0.845) and 0.702 (95% CI, 0.568–0.818), respectively.

We next developed a tripartite PD-1/TIGIT-based risk classification using ROC-derived cut-offs for CD4⁺PD-1−TIGIT⁺ and CD4⁺PD-1⁺TIGIT− T cells. We categorized patients with CD4⁺PD-1−TIGIT⁺ frequencies < 6.71% and CD4⁺PD-1⁺TIGIT− frequencies > 11.65% of total CD4⁺ T cells as low risk, and those with CD4⁺PD-1−TIGIT⁺ > 6.71% and CD4⁺PD-1⁺TIGIT− < 11.65% as high risk. We assigned all remaining patients—those with either both values above or both values below the cut-offs—to the intermediate-risk group (Figure 3E). Off-ECMO mortality rates were 10.53%, 16.22%, and 70.59% in the low-, intermediate-, and high-risk groups, respectively (p < 0.001; Figure 3F), demonstrating a markedly increased risk of off-ECMO mortality when elevated CD4⁺PD-1−TIGIT⁺ and reduced CD4⁺PD-1⁺TIGIT− T cell frequencies occurred together. Additional exploratory logistic regression analyses among successfully weaned patients showed that the PD-1/TIGIT-based risk classification remained associated with off-ECMO mortality after adjustment for selected clinical covariates (Supplementary Table 2). Given the limited number of off-ECMO mortality events, these findings should be interpreted as exploratory.

Discussion

The high mortality rate after successful VA-ECMO weaning indicates that key pathophysiological processes remain unresolved after decannulation.38 Several clinical models and predictors have been proposed to predict successful weaning from VA-ECMO, whereas their ability to predict post-weaning (off-ECMO) mortality remains limited.39 In this study, dynamic analyses combined with bootstrap validation identified two circulating CD4⁺ T cell subsets - CD4⁺PD-1−TIGIT⁺ and CD4⁺PD-1⁺TIGIT− cells - that were associated with off-ECMO mortality. Importantly, the combined shifts in these two subsets discriminated off-ECMO mortality more effectively than either subset alone. Compared with previous VA-ECMO prognostic models that mainly rely on pre-ECMO clinical and hemodynamic variables and prior immune studies that focused largely on cytokines and innate immune cell alterations,9–11,15–20 this PD-1/TIGIT-based CD4⁺ T cell signature may provide complementary adaptive immune information for early post-weaning risk stratification in VA-ECMO-supported patients. In exploratory multivariable analyses, the PD-1/TIGIT-based risk classification remained associated with off-ECMO mortality after adjustment for selected covariates, further supporting its potential clinical relevance.

In patients receiving ECMO, immune abnormalities, including systemic inflammatory activation, neutrophil–lymphocyte imbalance, monocyte dysfunction, and impaired T cell responses, may have been triggered by critical illnesses such as cardiogenic shock and sepsis before ECMO initiation.14 During ECMO support, exposure to the extracorporeal circuit and ischemia–reperfusion injury further activates a new wave of innate immune responses.14,24 However, many early innate immune signals diminish over time,40–43 whereas off-ECMO mortality is often driven by adverse events occurring after decannulation,44,45 limiting the predictive value of early innate markers. In contrast, adaptive immune suppression - particularly T cell dysfunction - develops over a longer period.46,47 Sustained inflammation, cytokine imbalance, and expansion of myeloid-derived suppressor cells (MDSCs) can jointly disrupt T cell homeostasis and impair T cell function.21–24,27 Consistent with prior human and animal studies,25,48 we observed rapid lymphopenia, an increased CD4/CD8 ratio, and redistribution of T cell compartments in patients supported with VA-ECMO. However, these changes likely reflect shared consequences of critical illness, ischemia–reperfusion injury, and ECMO exposure, as we found no associations between these parameters and clinical outcomes. Importantly, this redistribution did not appear to result from excessive T cell activation or increased thymic output, because frequencies of CD38⁺HLA-DR⁺ T cells and CD31⁺CD4⁺ TN did not increase in ECMO-supported patients (unpublished data). Similar immune features have been reported in sepsis, where persistent T cell lymphopenia - largely driven by lymphocyte apoptosis - associates with secondary infections, prolonged immunosuppression, and increased mortality.46 In addition, interleukin-7 (IL-7), a key cytokine for T cell survival that correlates with improved outcomes in critically ill patients,49 was largely undetectable in the plasma of VA-ECMO patients,27 which may further contribute to sustained T cell depletion. Nevertheless, likely due to the limited cohort size, these global T cell alterations did not differ across outcome groups and did not predict prognosis in our study.

The frequencies of CD4⁺PD-1−TIGIT⁺ and CD4⁺PD-1⁺TIGIT− T cells showed opposite associations with off-ECMO mortality, with higher CD4⁺PD-1−TIGIT⁺ frequencies and lower CD4⁺PD-1⁺TIGIT− frequencies in patients who died after weaning. The mechanisms underlying these divergent associations remain unclear. One possible explanation involves differences in differentiation status: in ECMO-supported patients, CD4⁺PD-1−TIGIT⁺ cells were enriched in both TCM and TEM compartments, whereas CD4⁺PD-1⁺TIGIT− cells were predominantly enriched in the TEM compartment. In general, CD4⁺ TCM cells are long-lived and lymphoid-homing, while CD4⁺ TEM cells rapidly produce cytokines and mediate immediate effector responses.50 Beyond phenotypic distribution, these subsets may also differ functionally. In patients with viral infections and lymphoma, CD4⁺PD-1⁺TIGIT− cells produce higher levels of IFN-γ than CD4⁺PD-1−TIGIT⁺ cells.51,52 In addition, in a mouse model of acute kidney injury, ischemia increased TIGIT expression on renal CD4⁺ T cells, and TIGIT deficiency reduced kidney injury after ischemic insult,53 supporting a functional role for TIGIT-expressing CD4⁺ T cells under ischemia-related stress.

Although PD-1 and TIGIT are both co-inhibitory receptors linked to T cell exhaustion, they signal through distinct pathways and may exert different functional effects. Upon binding PD-L1, PD-1 suppresses T cell proliferation, cytokine production, and survival, which can promote chronic infection and tumor progression.54 While persistent antigen exposure and inflammation typically drive classical T cell exhaustion,55 PD-1 can also act as a physiological checkpoint during acute infection to limit immune-mediated tissue damage.56 In contrast, TIGIT is expressed on activated T cells, regulatory T cells, and NK cells and inhibits T cell responses by competing with the co-stimulatory receptor CD226 for shared ligands, CD155 and CD112.57–59 Consistent with this role, increased TIGIT expression on T cells has been associated with immune dysfunction and poor outcomes in sepsis.60,61 Within this context, the opposing shifts in CD4⁺PD-1−TIGIT⁺ and CD4⁺PD-1⁺TIGIT− subsets associated with off-ECMO mortality may reflect an imbalance within the CD4⁺ T cell compartment, a hypothesis that warrants further mechanistic investigation.

This study has several strengths and potential clinical implications. First, it focused on off-ECMO mortality after successful weaning, a clinically important but less well-characterized outcome than on-ECMO mortality or overall in-hospital mortality. Second, the proposed PD-1/TIGIT-defined CD4⁺ T-cell signature was assessed using peripheral blood samples early during ECMO support, providing an opportunity for early identification of patients who remain at high risk after decannulation. Third, this signature is based on flow cytometry, requires only peripheral blood samples and a limited marker panel, and may therefore be relatively less costly and more feasible than high-throughput omics platforms. These features may make this approach more feasible for timely peripheral immune assessment in routine practice, although its clinical implementation requires further validation in prospective studies.

Study Limitations and Future Directions

This study has several important limitations. First, the single-center, retrospective design limits population representativeness and may introduce selection bias. Second, although we applied bootstrap resampling to test the robustness of the association between the PD-1/TIGIT-based CD4⁺ T cell signature and off-ECMO mortality, the proposed risk stratification model requires external validation in independent cohorts to establish generalizability across diverse clinical settings, which is essential for clinical translation. Third, although exploratory multivariable logistic regression analyses were performed, only 22 off-ECMO mortality events were observed. Therefore, overfitting, residual confounding, and model instability cannot be excluded, and the adjusted associations require validation in larger prospective cohorts. Fourth, we did not perform mechanistic studies of the CD4⁺PD-1−TIGIT⁺ and CD4⁺PD-1⁺TIGIT− T cell subsets, and therefore the biological roles of these subsets in the ECMO setting remain unclear.

Future studies should address these limitations from both clinical and mechanistic perspectives. Prospective multicenter cohorts are needed to validate the proposed cut-offs and risk classification, evaluate the incremental value of PD-1/TIGIT-defined CD4⁺ T cell subsets over established clinical scores and routine clinical variables, and standardize flow cytometry panels, gating strategies, and reporting procedures. Because off-ECMO mortality most commonly resulted from multiple organ failure, infection, or neurological injury, future studies should integrate T cell–based biomarkers with post-ECMO complication profiles to improve identification of high-risk patients. In parallel, single-cell and functional studies are needed to clarify whether these subsets directly contribute to post-ECMO immune dysfunction or mainly serve as biomarkers of post-weaning risk.

Conclusions

In summary, we identified an early PD-1/TIGIT-defined CD4⁺ T cell signature associated with off-ECMO mortality after VA-ECMO weaning, characterized by increased CD4⁺PD-1−TIGIT⁺ and decreased CD4⁺PD-1⁺TIGIT− T cells. This signature may provide complementary immune biomarker information for post-weaning risk stratification.